Dataset Viewer
model_id
stringlengths 8
65
| model_card
stringlengths 0
15.7k
| model_labels
listlengths |
|---|---|---|
Salesforce/blip-image-captioning-large |
# BLIP: Bootstrapping Language-Image Pre-training for Unified Vision-Language Understanding and Generation
Model card for image captioning pretrained on COCO dataset - base architecture (with ViT large backbone).
|  |
|:--:|
| <b> Pull figure from BLIP official repo | Image source: https://github.com/salesforce/BLIP </b>|
## TL;DR
Authors from the [paper](https://arxiv.org/abs/2201.12086) write in the abstract:
*Vision-Language Pre-training (VLP) has advanced the performance for many vision-language tasks. However, most existing pre-trained models only excel in either understanding-based tasks or generation-based tasks. Furthermore, performance improvement has been largely achieved by scaling up the dataset with noisy image-text pairs collected from the web, which is a suboptimal source of supervision. In this paper, we propose BLIP, a new VLP framework which transfers flexibly to both vision-language understanding and generation tasks. BLIP effectively utilizes the noisy web data by bootstrapping the captions, where a captioner generates synthetic captions and a filter removes the noisy ones. We achieve state-of-the-art results on a wide range of vision-language tasks, such as image-text retrieval (+2.7% in average recall@1), image captioning (+2.8% in CIDEr), and VQA (+1.6% in VQA score). BLIP also demonstrates strong generalization ability when directly transferred to videolanguage tasks in a zero-shot manner. Code, models, and datasets are released.*
## Usage
You can use this model for conditional and un-conditional image captioning
### Using the Pytorch model
#### Running the model on CPU
<details>
<summary> Click to expand </summary>
```python
import requests
from PIL import Image
from transformers import BlipProcessor, BlipForConditionalGeneration
processor = BlipProcessor.from_pretrained("Salesforce/blip-image-captioning-large")
model = BlipForConditionalGeneration.from_pretrained("Salesforce/blip-image-captioning-large")
img_url = 'https://storage.googleapis.com/sfr-vision-language-research/BLIP/demo.jpg'
raw_image = Image.open(requests.get(img_url, stream=True).raw).convert('RGB')
# conditional image captioning
text = "a photography of"
inputs = processor(raw_image, text, return_tensors="pt")
out = model.generate(**inputs)
print(processor.decode(out[0], skip_special_tokens=True))
# unconditional image captioning
inputs = processor(raw_image, return_tensors="pt")
out = model.generate(**inputs)
print(processor.decode(out[0], skip_special_tokens=True))
```
</details>
#### Running the model on GPU
##### In full precision
<details>
<summary> Click to expand </summary>
```python
import requests
from PIL import Image
from transformers import BlipProcessor, BlipForConditionalGeneration
processor = BlipProcessor.from_pretrained("Salesforce/blip-image-captioning-large")
model = BlipForConditionalGeneration.from_pretrained("Salesforce/blip-image-captioning-large").to("cuda")
img_url = 'https://storage.googleapis.com/sfr-vision-language-research/BLIP/demo.jpg'
raw_image = Image.open(requests.get(img_url, stream=True).raw).convert('RGB')
# conditional image captioning
text = "a photography of"
inputs = processor(raw_image, text, return_tensors="pt").to("cuda")
out = model.generate(**inputs)
print(processor.decode(out[0], skip_special_tokens=True))
# unconditional image captioning
inputs = processor(raw_image, return_tensors="pt").to("cuda")
out = model.generate(**inputs)
print(processor.decode(out[0], skip_special_tokens=True))
```
</details>
##### In half precision (`float16`)
<details>
<summary> Click to expand </summary>
```python
import torch
import requests
from PIL import Image
from transformers import BlipProcessor, BlipForConditionalGeneration
processor = BlipProcessor.from_pretrained("Salesforce/blip-image-captioning-large")
model = BlipForConditionalGeneration.from_pretrained("Salesforce/blip-image-captioning-large", torch_dtype=torch.float16).to("cuda")
img_url = 'https://storage.googleapis.com/sfr-vision-language-research/BLIP/demo.jpg'
raw_image = Image.open(requests.get(img_url, stream=True).raw).convert('RGB')
# conditional image captioning
text = "a photography of"
inputs = processor(raw_image, text, return_tensors="pt").to("cuda", torch.float16)
out = model.generate(**inputs)
print(processor.decode(out[0], skip_special_tokens=True))
# >>> a photography of a woman and her dog
# unconditional image captioning
inputs = processor(raw_image, return_tensors="pt").to("cuda", torch.float16)
out = model.generate(**inputs)
print(processor.decode(out[0], skip_special_tokens=True))
>>> a woman sitting on the beach with her dog
```
</details>
## Ethical Considerations
This release is for research purposes only in support of an academic paper. Our models, datasets, and code are not specifically designed or evaluated for all downstream purposes. We strongly recommend users evaluate and address potential concerns related to accuracy, safety, and fairness before deploying this model. We encourage users to consider the common limitations of AI, comply with applicable laws, and leverage best practices when selecting use cases, particularly for high-risk scenarios where errors or misuse could significantly impact people’s lives, rights, or safety. For further guidance on use cases, refer to our AUP and AI AUP.
## BibTex and citation info
```
@misc{https://doi.org/10.48550/arxiv.2201.12086,
doi = {10.48550/ARXIV.2201.12086},
url = {https://arxiv.org/abs/2201.12086},
author = {Li, Junnan and Li, Dongxu and Xiong, Caiming and Hoi, Steven},
keywords = {Computer Vision and Pattern Recognition (cs.CV), FOS: Computer and information sciences, FOS: Computer and information sciences},
title = {BLIP: Bootstrapping Language-Image Pre-training for Unified Vision-Language Understanding and Generation},
publisher = {arXiv},
year = {2022},
copyright = {Creative Commons Attribution 4.0 International}
}
``` | null |
Salesforce/blip-image-captioning-base |
# BLIP: Bootstrapping Language-Image Pre-training for Unified Vision-Language Understanding and Generation
Model card for image captioning pretrained on COCO dataset - base architecture (with ViT base backbone).
|  |
|:--:|
| <b> Pull figure from BLIP official repo | Image source: https://github.com/salesforce/BLIP </b>|
## TL;DR
Authors from the [paper](https://arxiv.org/abs/2201.12086) write in the abstract:
*Vision-Language Pre-training (VLP) has advanced the performance for many vision-language tasks. However, most existing pre-trained models only excel in either understanding-based tasks or generation-based tasks. Furthermore, performance improvement has been largely achieved by scaling up the dataset with noisy image-text pairs collected from the web, which is a suboptimal source of supervision. In this paper, we propose BLIP, a new VLP framework which transfers flexibly to both vision-language understanding and generation tasks. BLIP effectively utilizes the noisy web data by bootstrapping the captions, where a captioner generates synthetic captions and a filter removes the noisy ones. We achieve state-of-the-art results on a wide range of vision-language tasks, such as image-text retrieval (+2.7% in average recall@1), image captioning (+2.8% in CIDEr), and VQA (+1.6% in VQA score). BLIP also demonstrates strong generalization ability when directly transferred to videolanguage tasks in a zero-shot manner. Code, models, and datasets are released.*
## Usage
You can use this model for conditional and un-conditional image captioning
### Using the Pytorch model
#### Running the model on CPU
<details>
<summary> Click to expand </summary>
```python
import requests
from PIL import Image
from transformers import BlipProcessor, BlipForConditionalGeneration
processor = BlipProcessor.from_pretrained("Salesforce/blip-image-captioning-base")
model = BlipForConditionalGeneration.from_pretrained("Salesforce/blip-image-captioning-base")
img_url = 'https://storage.googleapis.com/sfr-vision-language-research/BLIP/demo.jpg'
raw_image = Image.open(requests.get(img_url, stream=True).raw).convert('RGB')
# conditional image captioning
text = "a photography of"
inputs = processor(raw_image, text, return_tensors="pt")
out = model.generate(**inputs)
print(processor.decode(out[0], skip_special_tokens=True))
# >>> a photography of a woman and her dog
# unconditional image captioning
inputs = processor(raw_image, return_tensors="pt")
out = model.generate(**inputs)
print(processor.decode(out[0], skip_special_tokens=True))
>>> a woman sitting on the beach with her dog
```
</details>
#### Running the model on GPU
##### In full precision
<details>
<summary> Click to expand </summary>
```python
import requests
from PIL import Image
from transformers import BlipProcessor, BlipForConditionalGeneration
processor = BlipProcessor.from_pretrained("Salesforce/blip-image-captioning-base")
model = BlipForConditionalGeneration.from_pretrained("Salesforce/blip-image-captioning-base").to("cuda")
img_url = 'https://storage.googleapis.com/sfr-vision-language-research/BLIP/demo.jpg'
raw_image = Image.open(requests.get(img_url, stream=True).raw).convert('RGB')
# conditional image captioning
text = "a photography of"
inputs = processor(raw_image, text, return_tensors="pt").to("cuda")
out = model.generate(**inputs)
print(processor.decode(out[0], skip_special_tokens=True))
# >>> a photography of a woman and her dog
# unconditional image captioning
inputs = processor(raw_image, return_tensors="pt").to("cuda")
out = model.generate(**inputs)
print(processor.decode(out[0], skip_special_tokens=True))
>>> a woman sitting on the beach with her dog
```
</details>
##### In half precision (`float16`)
<details>
<summary> Click to expand </summary>
```python
import torch
import requests
from PIL import Image
from transformers import BlipProcessor, BlipForConditionalGeneration
processor = BlipProcessor.from_pretrained("Salesforce/blip-image-captioning-base")
model = BlipForConditionalGeneration.from_pretrained("Salesforce/blip-image-captioning-base", torch_dtype=torch.float16).to("cuda")
img_url = 'https://storage.googleapis.com/sfr-vision-language-research/BLIP/demo.jpg'
raw_image = Image.open(requests.get(img_url, stream=True).raw).convert('RGB')
# conditional image captioning
text = "a photography of"
inputs = processor(raw_image, text, return_tensors="pt").to("cuda", torch.float16)
out = model.generate(**inputs)
print(processor.decode(out[0], skip_special_tokens=True))
# >>> a photography of a woman and her dog
# unconditional image captioning
inputs = processor(raw_image, return_tensors="pt").to("cuda", torch.float16)
out = model.generate(**inputs)
print(processor.decode(out[0], skip_special_tokens=True))
>>> a woman sitting on the beach with her dog
```
</details>
## Ethical Considerations
This release is for research purposes only in support of an academic paper. Our models, datasets, and code are not specifically designed or evaluated for all downstream purposes. We strongly recommend users evaluate and address potential concerns related to accuracy, safety, and fairness before deploying this model. We encourage users to consider the common limitations of AI, comply with applicable laws, and leverage best practices when selecting use cases, particularly for high-risk scenarios where errors or misuse could significantly impact people’s lives, rights, or safety. For further guidance on use cases, refer to our AUP and AI AUP.
## BibTex and citation info
```
@misc{https://doi.org/10.48550/arxiv.2201.12086,
doi = {10.48550/ARXIV.2201.12086},
url = {https://arxiv.org/abs/2201.12086},
author = {Li, Junnan and Li, Dongxu and Xiong, Caiming and Hoi, Steven},
keywords = {Computer Vision and Pattern Recognition (cs.CV), FOS: Computer and information sciences, FOS: Computer and information sciences},
title = {BLIP: Bootstrapping Language-Image Pre-training for Unified Vision-Language Understanding and Generation},
publisher = {arXiv},
year = {2022},
copyright = {Creative Commons Attribution 4.0 International}
}
```
| null |
Salesforce/blip2-opt-2.7b |
# BLIP-2, OPT-2.7b, pre-trained only
BLIP-2 model, leveraging [OPT-2.7b](https://huggingface.co/facebook/opt-2.7b) (a large language model with 2.7 billion parameters).
It was introduced in the paper [BLIP-2: Bootstrapping Language-Image Pre-training with Frozen Image Encoders and Large Language Models](https://arxiv.org/abs/2301.12597) by Li et al. and first released in [this repository](https://github.com/salesforce/LAVIS/tree/main/projects/blip2).
Disclaimer: The team releasing BLIP-2 did not write a model card for this model so this model card has been written by the Hugging Face team.
## Model description
BLIP-2 consists of 3 models: a CLIP-like image encoder, a Querying Transformer (Q-Former) and a large language model.
The authors initialize the weights of the image encoder and large language model from pre-trained checkpoints and keep them frozen
while training the Querying Transformer, which is a BERT-like Transformer encoder that maps a set of "query tokens" to query embeddings,
which bridge the gap between the embedding space of the image encoder and the large language model.
The goal for the model is simply to predict the next text token, giving the query embeddings and the previous text.
<img src="https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/transformers/model_doc/blip2_architecture.jpg"
alt="drawing" width="600"/>
This allows the model to be used for tasks like:
- image captioning
- visual question answering (VQA)
- chat-like conversations by feeding the image and the previous conversation as prompt to the model
## Direct Use and Downstream Use
You can use the raw model for conditional text generation given an image and optional text. See the [model hub](https://huggingface.co/models?search=Salesforce/blip) to look for
fine-tuned versions on a task that interests you.
## Bias, Risks, Limitations, and Ethical Considerations
BLIP2-OPT uses off-the-shelf OPT as the language model. It inherits the same risks and limitations as mentioned in Meta's model card.
> Like other large language models for which the diversity (or lack thereof) of training
> data induces downstream impact on the quality of our model, OPT-175B has limitations in terms
> of bias and safety. OPT-175B can also have quality issues in terms of generation diversity and
> hallucination. In general, OPT-175B is not immune from the plethora of issues that plague modern
> large language models.
>
BLIP2 is fine-tuned on image-text datasets (e.g. [LAION](https://laion.ai/blog/laion-400-open-dataset/) ) collected from the internet. As a result the model itself is potentially vulnerable to generating equivalently inappropriate content or replicating inherent biases in the underlying data.
BLIP2 has not been tested in real world applications. It should not be directly deployed in any applications. Researchers should first carefully assess the safety and fairness of the model in relation to the specific context they’re being deployed within.
## Ethical Considerations
This release is for research purposes only in support of an academic paper. Our models, datasets, and code are not specifically designed or evaluated for all downstream purposes. We strongly recommend users evaluate and address potential concerns related to accuracy, safety, and fairness before deploying this model. We encourage users to consider the common limitations of AI, comply with applicable laws, and leverage best practices when selecting use cases, particularly for high-risk scenarios where errors or misuse could significantly impact people’s lives, rights, or safety. For further guidance on use cases, refer to our AUP and AI AUP.
### How to use
For code examples, we refer to the [documentation](https://huggingface.co/docs/transformers/main/en/model_doc/blip-2#transformers.Blip2ForConditionalGeneration.forward.example).
### Memory requirements
The memory requirements differ based on the precision one uses. One can use 4-bit inference using [Bitsandbytes](https://huggingface.co/blog/4bit-transformers-bitsandbytes), which greatly reduce the memory requirements.
| dtype | Largest Layer or Residual Group | Total Size | Training using Adam |
|-------------------|---------------------------------|------------|----------------------|
| float32 | 490.94 MB | 14.43 GB | 57.72 GB |
| float16/bfloat16 | 245.47 MB | 7.21 GB | 28.86 GB |
| int8 | 122.73 MB | 3.61 GB | 14.43 GB |
| int4 | 61.37 MB | 1.8 GB | 7.21 GB |
#### Running the model on CPU
<details>
<summary> Click to expand </summary>
```python
import requests
from PIL import Image
from transformers import Blip2Processor, Blip2ForConditionalGeneration
processor = Blip2Processor.from_pretrained("Salesforce/blip2-opt-2.7b")
model = Blip2ForConditionalGeneration.from_pretrained("Salesforce/blip2-opt-2.7b")
img_url = 'https://storage.googleapis.com/sfr-vision-language-research/BLIP/demo.jpg'
raw_image = Image.open(requests.get(img_url, stream=True).raw).convert('RGB')
question = "how many dogs are in the picture?"
inputs = processor(raw_image, question, return_tensors="pt")
out = model.generate(**inputs)
print(processor.decode(out[0], skip_special_tokens=True).strip())
```
</details>
#### Running the model on GPU
##### In full precision
<details>
<summary> Click to expand </summary>
```python
# pip install accelerate
import requests
from PIL import Image
from transformers import Blip2Processor, Blip2ForConditionalGeneration
processor = Blip2Processor.from_pretrained("Salesforce/blip2-opt-2.7b")
model = Blip2ForConditionalGeneration.from_pretrained("Salesforce/blip2-opt-2.7b", device_map="auto")
img_url = 'https://storage.googleapis.com/sfr-vision-language-research/BLIP/demo.jpg'
raw_image = Image.open(requests.get(img_url, stream=True).raw).convert('RGB')
question = "how many dogs are in the picture?"
inputs = processor(raw_image, question, return_tensors="pt").to("cuda")
out = model.generate(**inputs)
print(processor.decode(out[0], skip_special_tokens=True).strip())
```
</details>
##### In half precision (`float16`)
<details>
<summary> Click to expand </summary>
```python
# pip install accelerate
import torch
import requests
from PIL import Image
from transformers import Blip2Processor, Blip2ForConditionalGeneration
processor = Blip2Processor.from_pretrained("Salesforce/blip2-opt-2.7b")
model = Blip2ForConditionalGeneration.from_pretrained("Salesforce/blip2-opt-2.7b", torch_dtype=torch.float16, device_map="auto")
img_url = 'https://storage.googleapis.com/sfr-vision-language-research/BLIP/demo.jpg'
raw_image = Image.open(requests.get(img_url, stream=True).raw).convert('RGB')
question = "how many dogs are in the picture?"
inputs = processor(raw_image, question, return_tensors="pt").to("cuda", torch.float16)
out = model.generate(**inputs)
print(processor.decode(out[0], skip_special_tokens=True).strip())
```
</details>
##### In 8-bit precision (`int8`)
<details>
<summary> Click to expand </summary>
```python
# pip install accelerate bitsandbytes
import torch
import requests
from PIL import Image
from transformers import Blip2Processor, Blip2ForConditionalGeneration
processor = Blip2Processor.from_pretrained("Salesforce/blip2-opt-2.7b")
model = Blip2ForConditionalGeneration.from_pretrained("Salesforce/blip2-opt-2.7b", load_in_8bit=True, device_map="auto")
img_url = 'https://storage.googleapis.com/sfr-vision-language-research/BLIP/demo.jpg'
raw_image = Image.open(requests.get(img_url, stream=True).raw).convert('RGB')
question = "how many dogs are in the picture?"
inputs = processor(raw_image, question, return_tensors="pt").to("cuda", torch.float16)
out = model.generate(**inputs)
print(processor.decode(out[0], skip_special_tokens=True).strip())
```
</details> | null |
microsoft/git-base |
# GIT (GenerativeImage2Text), base-sized
GIT (short for GenerativeImage2Text) model, base-sized version. It was introduced in the paper [GIT: A Generative Image-to-text Transformer for Vision and Language](https://arxiv.org/abs/2205.14100) by Wang et al. and first released in [this repository](https://github.com/microsoft/GenerativeImage2Text).
Disclaimer: The team releasing GIT did not write a model card for this model so this model card has been written by the Hugging Face team.
## Model description
GIT is a Transformer decoder conditioned on both CLIP image tokens and text tokens. The model is trained using "teacher forcing" on a lot of (image, text) pairs.
The goal for the model is simply to predict the next text token, giving the image tokens and previous text tokens.
The model has full access to (i.e. a bidirectional attention mask is used for) the image patch tokens, but only has access to the previous text tokens (i.e. a causal attention mask is used for the text tokens) when predicting the next text token.
This allows the model to be used for tasks like:
- image and video captioning
- visual question answering (VQA) on images and videos
- even image classification (by simply conditioning the model on the image and asking it to generate a class for it in text).
## Intended uses & limitations
You can use the raw model for image captioning. See the [model hub](https://huggingface.co/models?search=microsoft/git) to look for
fine-tuned versions on a task that interests you.
### How to use
For code examples, we refer to the [documentation](https://huggingface.co/docs/transformers/main/model_doc/git#transformers.GitForCausalLM.forward.example).
## Training data
From the paper:
> We collect 0.8B image-text pairs for pre-training, which include COCO (Lin et al., 2014), Conceptual Captions
(CC3M) (Sharma et al., 2018), SBU (Ordonez et al., 2011), Visual Genome (VG) (Krishna et al., 2016),
Conceptual Captions (CC12M) (Changpinyo et al., 2021), ALT200M (Hu et al., 2021a), and an extra 0.6B
data following a similar collection procedure in Hu et al. (2021a).
=> however this is for the model referred to as "GIT" in the paper, which is not open-sourced.
This checkpoint is "GIT-base", which is a smaller variant of GIT trained on 10 million image-text pairs.
See table 11 in the [paper](https://arxiv.org/abs/2205.14100) for more details.
### Preprocessing
We refer to the original repo regarding details for preprocessing during training.
During validation, one resizes the shorter edge of each image, after which center cropping is performed to a fixed-size resolution. Next, frames are normalized across the RGB channels with the ImageNet mean and standard deviation.
## Evaluation results
For evaluation results, we refer readers to the [paper](https://arxiv.org/abs/2205.14100). | null |
Dataseeds/LLaVA-OneVision-Qwen2-0.5b-ov-DSD-FineTune |
# LLaVA-OneVision-Qwen2-0.5b Fine-tuned on DataSeeds.AI Dataset
This model is a LoRA (Low-Rank Adaptation) fine-tuned version of [lmms-lab/llava-onevision-qwen2-0.5b-ov](https://huggingface.co/lmms-lab/llava-onevision-qwen2-0.5b-ov) specialized for photography scene analysis and description generation. The model was presented in the paper [Peer-Ranked Precision: Creating a Foundational Dataset for Fine-Tuning Vision Models from DataSeeds' Annotated Imagery](https://huggingface.co/papers/2506.05673). The model was fine-tuned on the [DataSeeds.AI Sample Dataset (DSD)](https://huggingface.co/datasets/Dataseeds/DataSeeds.AI-Sample-Dataset-DSD) to enhance its capabilities in generating detailed, accurate descriptions of photographic content.
## Model Description
- **Base Model**: [LLaVA-OneVision-Qwen2-0.5b](https://huggingface.co/lmms-lab/llava-onevision-qwen2-0.5b-ov)
- **Vision Encoder**: [SigLIP-SO400M-patch14-384](https://huggingface.co/google/siglip-so400m-patch14-384)
- **Language Model**: Qwen2-0.5B (896M parameters)
- **Fine-tuning Method**: LoRA (Low-Rank Adaptation) with PEFT
- **Total Parameters**: ~917M (513M trainable during fine-tuning, 56% of total)
- **Multimodal Projector**: 1.84M parameters (100% trainable)
- **Precision**: BFloat16
- **Task**: Photography scene analysis and detailed image description
### LoRA Configuration
- **LoRA Rank (r)**: 32
- **LoRA Alpha**: 32
- **LoRA Dropout**: 0.1
- **Target Modules**: `v_proj`, `k_proj`, `q_proj`, `up_proj`, `gate_proj`, `down_proj`, `o_proj`
- **Tunable Components**: `mm_mlp_adapter`, `mm_language_model`
## Training Details
### Dataset
The model was fine-tuned on the DataSeeds.AI Sample Dataset, a curated collection of photography images with detailed scene descriptions focusing on:
- Compositional elements and camera perspectives
- Lighting conditions and visual ambiance
- Product identification and technical details
- Photographic style and mood analysis
### Training Configuration
| Parameter | Value |
|-----------|-------|
| **Learning Rate** | 1e-5 |
| **Optimizer** | AdamW |
| **Learning Rate Schedule** | Cosine decay |
| **Warmup Ratio** | 0.03 |
| **Weight Decay** | 0.01 |
| **Batch Size** | 2 |
| **Gradient Accumulation Steps** | 8 (effective batch size: 16) |
| **Training Epochs** | 3 |
| **Max Sequence Length** | 8192 |
| **Max Gradient Norm** | 0.5 |
| **Precision** | BFloat16 |
| **Hardware** | Single NVIDIA A100 40GB |
| **Training Time** | 30 hours |
### Training Strategy
- **Validation Frequency**: Every 50 steps for precise checkpoint selection
- **Best Checkpoint**: Step 1,750 (epoch 2.9) with validation loss of 1.83
- **Mixed Precision**: BFloat16 with gradient checkpointing for memory efficiency
- **System Prompt**: Consistent template requesting scene descriptions across all samples
## Performance
### Quantitative Results
The fine-tuned model shows significant improvements across all evaluation metrics compared to the base model:
| Metric | Base Model | Fine-tuned | Absolute Δ | Relative Δ |
|--------|------------|------------|------------|------------|
| **BLEU-4** | 0.0199 | **0.0246** | +0.0048 | **+24.09%** |
| **ROUGE-L** | 0.2089 | **0.2140** | +0.0051 | **+2.44%** |
| **BERTScore F1** | 0.2751 | **0.2789** | +0.0039 | **+1.40%** |
| **CLIPScore** | 0.3247 | **0.3260** | +0.0013 | **+0.41%** |
### Key Improvements
- **Enhanced N-gram Precision**: 24% improvement in BLEU-4 indicates significantly better word sequence accuracy
- **Better Sequential Information**: ROUGE-L improvement shows enhanced capture of longer matching sequences
- **Improved Semantic Understanding**: BERTScore gains demonstrate better contextual relationships
- **Maintained Visual-Semantic Alignment**: CLIPScore preservation with slight improvement
### Inference Performance
- **Processing Speed**: 2.30 seconds per image (NVIDIA A100 40GB)
- **Memory Requirements**: Optimized for single GPU inference
## Usage
### Installation
```bash
pip install transformers torch peft pillow
```
### Basic Usage
```python
from peft import PeftModel
from transformers import AutoModelForCausalLM, AutoTokenizer, AutoProcessor
import torch
from PIL import Image
# Load base model and processor
base_model = AutoModelForCausalLM.from_pretrained(
"lmms-lab/llava-onevision-qwen2-0.5b-ov",
torch_dtype=torch.bfloat16,
device_map="auto",
trust_remote_code=True
)
processor = AutoProcessor.from_pretrained("lmms-lab/llava-onevision-qwen2-0.5b-ov")
# Load LoRA adapter
model = PeftModel.from_pretrained(
base_model,
"Dataseeds/LLaVA-OneVision-Qwen2-0.5b-ov-DSD-FineTune"
)
# Load and process image
image = Image.open("your_image.jpg")
prompt = "Describe this image in detail, focusing on the composition, lighting, and visual elements."
inputs = processor(prompt, image, return_tensors="pt").to(model.device)
# Generate description
with torch.no_grad():
outputs = model.generate(
**inputs,
max_new_tokens=512,
do_sample=True,
temperature=0.7,
top_p=0.9
)
description = processor.decode(outputs[0], skip_special_tokens=True)
print(description)
```
### Advanced Usage with Custom Prompts
```python
# Photography-specific prompts that work well with this model
prompts = [
"Analyze the photographic composition and lighting in this image.",
"Describe the technical aspects and visual mood of this photograph.",
"Provide a detailed scene description focusing on the subject and environment."
]
for prompt in prompts:
inputs = processor(prompt, image, return_tensors="pt").to(model.device)
outputs = model.generate(**inputs, max_new_tokens=256, temperature=0.7)
description = processor.decode(outputs[0], skip_special_tokens=True)
print(f"Prompt: {prompt}")
print(f"Description: {description}
")
```
## Model Architecture
The model maintains the LLaVA-OneVision architecture with the following components:
- **Vision Encoder**: SigLIP-SO400M with hierarchical feature extraction
- **Language Model**: Qwen2-0.5B with 24 layers, 14 attention heads
- **Multimodal Projector**: 2-layer MLP with GELU activation (mlp2x_gelu)
- **Image Processing**: Supports "anyres_max_9" aspect ratio with dynamic grid pinpoints
- **Context Length**: 32,768 tokens with sliding window attention
### Technical Specifications
- **Hidden Size**: 896
- **Intermediate Size**: 4,864
- **Attention Heads**: 14 (2 key-value heads)
- **RMS Norm Epsilon**: 1e-6
- **RoPE Theta**: 1,000,000
- **Image Token Index**: 151646
- **Max Image Grid**: Up to 2304×2304 pixels with dynamic tiling
## Training Data
The DataSeeds.AI Sample Dataset contains curated photography images with comprehensive annotations including:
- **Scene Descriptions**: Detailed textual descriptions of visual content
- **Technical Metadata**: Camera settings, composition details
- **Style Analysis**: Photographic techniques and artistic elements
- **Quality Annotations**: Professional photography standards
The dataset focuses on enhancing the model's ability to:
- Identify specific products and technical details accurately
- Describe lighting conditions and photographic ambiance
- Analyze compositional elements and camera perspectives
- Generate contextually aware scene descriptions
## Limitations and Considerations
### Model Limitations
- **Domain Specialization**: Optimized for photography; may have reduced performance on general vision-language tasks
- **Base Model Inheritance**: Inherits limitations from LLaVA-OneVision base model
- **Training Data Bias**: May reflect biases present in the DataSeeds.AI dataset
- **Language Support**: Primarily trained and evaluated on English descriptions
### Recommended Use Cases
- ✅ Photography scene analysis and description
- ✅ Product photography captioning
- ✅ Technical photography analysis
- ✅ Visual content generation for photography applications
- ⚠️ General-purpose vision-language tasks (may have reduced performance)
- ❌ Non-photographic image analysis (not optimized for this use case)
### Ethical Considerations
- The model may perpetuate biases present in photography datasets
- Generated descriptions should be reviewed for accuracy in critical applications
- Consider potential cultural biases in photographic style interpretation
## Citation
If you use this model in your research or applications, please cite:
```bibtex
@article{abdoli2025peerranked,
title={Peer-Ranked Precision: Creating a Foundational Dataset for Fine-Tuning Vision Models from DataSeeds' Annotated Imagery},
author={Sajjad Abdoli and Freeman Lewin and Gediminas Vasiliauskas and Fabian Schonholz},
journal={arXiv preprint arXiv:2506.05673},
year={2025},
}
@misc{llava-onevision-dsd-finetune-2024,
title={LLaVA-OneVision Fine-tuned on DataSeeds.AI Dataset for Photography Scene Analysis},
author={DataSeeds.AI},
year={2024},
publisher={Hugging Face},
url={https://huggingface.co/Dataseeds/LLaVA-OneVision-Qwen2-0.5b-ov-DSD-FineTune},
note={LoRA fine-tuned model for enhanced photography description generation}
}
@article{li2024llavaonevision,
title={LLaVA-OneVision: Easy Visual Task Transfer},
author={Li, Bo and Zhang, Yuanhan and Guo, Dong and Zhang, Renrui and Li, Feng and Zhang, Hao and Zhang, Kaichen and Liu, Yanwei and Wang, Ziwei and Gao, Peng},
journal={arXiv preprint arXiv:2408.03326},
year={2024}
}
@article{hu2022lora,
title={LoRA: Low-Rank Adaptation of Large Language Models},
author={Hu, Edward J and Shen, Yelong and Wallis, Phillip and Allen-Zhu, Zeyuan and Li, Yuanzhi and Wang, Shean and Wang, Lu and Chen, Weizhu},
journal={arXiv preprint arXiv:2106.09685},
year={2021}
}
```
## License
This model is released under the Apache 2.0 license, consistent with the base LLaVA-OneVision model licensing terms.
## Acknowledgments
- **Base Model**: Thanks to LMMS Lab for the LLaVA-OneVision model
- **Vision Encoder**: Thanks to Google Research for the SigLIP model
- **Dataset**: GuruShots photography community for the source imagery
- **Framework**: Hugging Face PEFT library for efficient fine-tuning capabilities
---
*For questions, issues, or collaboration opportunities, please visit the [model repository](https://huggingface.co/Dataseeds/LLaVA-OneVision-Qwen2-0.5b-ov-DSD-FineTune) or contact the DataSeeds.AI team.* | null |
Dataseeds/BLIP2-opt-2.7b-DSD-FineTune |
# BLIP2-OPT-2.7B Fine-tuned on DataSeeds.AI Dataset
Code: https://github.com/DataSeeds-ai/DSD-finetune-blip-llava
This model is a fine-tuned version of [Salesforce/blip2-opt-2.7b](https://huggingface.co/Salesforce/blip2-opt-2.7b) specialized for photography scene analysis and technical description generation. The model was fine-tuned on the [DataSeeds.AI Sample Dataset (DSD)](https://huggingface.co/datasets/Dataseeds/DataSeeds-Sample-Dataset-DSD) to enhance its capabilities in generating detailed photographic descriptions with focus on composition, lighting, and technical aspects.
The model was presented in the paper [Peer-Ranked Precision: Creating a Foundational Dataset for Fine-Tuning Vision Models from DataSeeds' Annotated Imagery](https://huggingface.co/papers/2506.05673).
## Model Description
- **Base Model**: [BLIP2-OPT-2.7B](https://huggingface.co/Salesforce/blip2-opt-2.7b)
- **Vision Encoder**: EVA-CLIP ViT-g/14
- **Language Model**: OPT-2.7B (2.7 billion parameters)
- **Architecture**: BLIP-2 with Q-Former bridging vision and language
- **Fine-tuning Approach**: Full model fine-tuning (ViT unfrozen)
- **Task**: Photography scene analysis and detailed image captioning
- **Precision**: Mixed Precision (AMP enabled)
- **Image Resolution**: 364×364 pixels
## Training Details
### Dataset
The model was fine-tuned on the DataSeeds.AI Sample Dataset (DSD), containing 10,610 curated photography images with comprehensive annotations:
- **Training Set**: 9,549 image-text pairs (90%)
- **Validation Set**: 1,061 image-text pairs (10%)
- **Content Focus**: Technical scene descriptions, compositional analysis, lighting conditions
- **Annotation Quality**: Professional photography standards with 15+ word descriptions
### Training Configuration
| Parameter | Value |
|-----------|-------|
| **Learning Rate** | 1e-5 |
| **Optimizer** | AdamW |
| **LR Schedule** | Linear warmup + Cosine decay |
| **Warmup Steps** | 17 |
| **Weight Decay** | 0.01 |
| **Batch Size** | 8 |
| **Gradient Accumulation** | 1 |
| **Training Epochs** | 10 |
| **Mixed Precision** | AMP enabled |
| **Hardware** | Single NVIDIA A100 80GB |
| **Vision Encoder** | Unfrozen (trainable) |
| **Gradient Checkpointing** | Enabled |
### Generation Configuration
| Parameter | Value |
|-----------|-------|
| **Max Length** | 100 tokens |
| **Min Length** | 8 tokens |
| **Num Beams** | 5 |
| **Beam Search** | Enabled |
| **Task** | Captioning |
### Checkpoint Selection
- **Selection Metric**: Aggregate validation score (CIDEr + BLEU-4)
- **Best Epoch**: Epoch 1 (aggregate score: 0.2626)
- **Training Loss**: Decreased from 2.780 (epoch 1) to 1.692 (epoch 10)
- **Final Model**: `checkpoint_best.pth` (selected based on validation performance)
## Performance
### Quantitative Results
The fine-tuned model shows significant improvements in lexical overlap metrics, with notable trade-offs in semantic understanding:
| Metric | Base Model | Fine-tuned | Absolute Δ | Relative Δ |
|--------|------------|------------|------------|------------|
| **BLEU-4** | 0.001 | **0.047** | +0.046 | **+4600%*** |
| **ROUGE-L** | 0.126 | **0.242** | +0.116 | **+92.06%** |
| **BERTScore F1** | 0.0545 | **-0.0537** | -0.1082 | **-198.53%** |
| **CLIPScore** | 0.2854 | **0.2583** | -0.0271 | **-9.49%** |
*_Note: The extreme BLEU-4 improvement is due to the very low baseline (0.001), making relative improvements appear dramatic._
### Key Observations
**Strengths:**
- **Enhanced Lexical Matching**: Substantial improvements in BLEU-4 and ROUGE-L indicate better n-gram alignment with reference descriptions
- **Photography Terminology**: Model learned to incorporate photographic vocabulary and technical terms
- **Compositional Awareness**: Improved description of camera angles, lighting conditions, and visual elements
**Trade-offs:**
- **Semantic Coherence**: Negative BERTScore suggests divergence from reference semantic patterns
- **Visual-Text Alignment**: Moderate decrease in CLIPScore indicates reduced image-text semantic alignment
- **Repetitive Patterns**: Tendency toward formulaic descriptions with some repetition
### Performance Characteristics
The fine-tuning results reveal a model that has specialized in lexical pattern matching for photography descriptions but with trade-offs in semantic understanding. This suggests the model is particularly suited for applications requiring technical photography terminology rather than general-purpose image captioning.
## Usage
### Installation
```bash
pip install transformers torch pillow
```
### Basic Usage
```python
from transformers import Blip2Processor, Blip2ForConditionalGeneration
import torch
from PIL import Image
# Load model and processor
processor = Blip2Processor.from_pretrained("Dataseeds/BLIP2-opt-2.7b-DSD-FineTune")
model = Blip2ForConditionalGeneration.from_pretrained(
"Dataseeds/BLIP2-opt-2.7b-DSD-FineTune",
torch_dtype=torch.float16,
device_map="auto"
)
# Load and process image
image = Image.open("your_image.jpg")
inputs = processor(image, return_tensors="pt").to(model.device, torch.float16)
# Generate caption
generated_ids = model.generate(
**inputs,
max_length=100,
min_length=8,
num_beams=5,
do_sample=False
)
caption = processor.decode(generated_ids[0], skip_special_tokens=True)
print(f"Generated caption: {caption}")
```
## Model Architecture
The model maintains the BLIP-2 architecture with the following components:
### Core Architecture
- **Vision Encoder**: EVA-CLIP ViT-g/14 (unfrozen during fine-tuning)
- **Q-Former**: 32-layer transformer bridging vision and language modalities
- **Language Model**: OPT-2.7B (2.7 billion parameters)
- **Architecture**: BLIP-2 with Q-Former bridging vision and language
- **Bootstrapping**: Two-stage pre-training methodology preserved
### Technical Specifications
- **Vision Resolution**: 364×364 pixels
- **Vision Patch Size**: 14×14
- **Q-Former Queries**: 32 learnable queries
- **Language Model Layers**: 32
- **Total Parameters**: ~2.7B (language model) + vision components
- **Precision**: Mixed precision (FP16/FP32)
### Fine-tuning Approach
- **Vision Encoder**: Trainable (freeze_vit: False)
- **Q-Former**: Trainable
- **Language Model**: Trainable
- **Full Model**: End-to-end fine-tuning enabled
- **Gradient Checkpointing**: Memory optimization enabled
## Training Data & Methodology
### Dataset Characteristics
The DataSeeds.AI Sample Dataset (DSD) focuses on:
- **Technical Photography**: Camera settings, composition analysis
- **Lighting Descriptions**: Ambient, directional, studio lighting analysis
- **Subject Matter**: Diverse photographic subjects and styles
- **Annotation Style**: Technical scene descriptions (20-30 words typical)
### Data Processing
- **Image Preprocessing**: `blip2_image_train` (364×364 resolution)
- **Text Processing**: `blip_caption` processor
- **Evaluation**: `blip_image_eval` for consistent validation
- **Format**: Input-output pairs with scene analysis prompts
### Recommended Use Cases
- ✅ Photography scene analysis and technical descriptions
- ✅ Camera composition and lighting analysis
- ✅ Product photography captioning
- ✅ Photography education and training applications
- ⚠️ General-purpose image captioning (may show repetitive patterns)
- ❌ Non-photographic content analysis (not optimized)
## Citation
If you use this model in your research or applications, please cite:
```bibtex
@article{abdoli2025peerranked,
title={Peer-Ranked Precision: Creating a Foundational Dataset for Fine-Tuning Vision Models from GuruShots' Annotated Imagery},
author={Sajjad Abdoli and Freeman Lewin and Gediminas Vasiliauskas and Fabian Schonholz},
journal={arXiv preprint arXiv:2506.05673},
year={2025},
}
@misc{blip2-opt-dsd-finetune-2024,
title={BLIP2-OPT-2.7B Fine-tuned on DataSeeds.AI Dataset for Photography Analysis},
author={Dataseeds},
year={2024},
publisher={Hugging Face},
url={https://huggingface.co/Dataseeds/BLIP2-opt-2.7b-DSD-FineTune},
note={Fine-tuned model for photography scene analysis and technical description}
}
@inproceedings{li2023blip2,
title={BLIP-2: Bootstrapping Language-Image Pre-training with Frozen Image Encoders and Large Language Models},
author={Li, Junnan and Li, Dongxu and Xiong, Caiming and Hoi, Steven},
booktitle={International Conference on Machine Learning},
pages={19730--19742},
year={2023},
organization={PMLR}
}
@article{zhang2022opt,
title={OPT: Open Pre-trained Transformer Language Models},
author={Zhang, Susan and Roller, Stephen and Goyal, Naman and Artetxe, Mikel and Chen, Moya and Chen, Shuohui and Dewan, Christopher and Diab, Mona and Li, Xian and Lin, Xi Victoria and others},
journal={arXiv preprint arXiv:2205.01068},
year={2022}
}
```
## License
This model is released under the MIT license, consistent with the base BLIP2 model licensing terms.
## Acknowledgments
- **Base Model**: Salesforce Research for BLIP2 architecture and pre-training
- **Language Model**: Meta AI for OPT-2.7B foundation
- **Vision Encoder**: OpenAI for EVA-CLIP vision components
- **Dataset**: GuruShots photography community for source imagery
- **Framework**: Hugging Face Transformers for model infrastructure
## Training Artifacts
This repository includes comprehensive training artifacts:
- **Checkpoints**: All epoch checkpoints (0-9) plus best performing checkpoint
- **Evaluation Results**: Validation metrics for each epoch
- **Training Logs**: Complete training and evaluation logs
- **Configuration**: Original training configuration and hyperparameters
---
*For questions, issues, or collaboration opportunities, please visit the [model repository](https://huggingface.co/Dataseeds/BLIP2-opt-2.7b-DSD-FineTune) or contact the DataSeeds.AI team.* | null |
nlpconnect/vit-gpt2-image-captioning |
# nlpconnect/vit-gpt2-image-captioning
This is an image captioning model trained by @ydshieh in [flax ](https://github.com/huggingface/transformers/tree/main/examples/flax/image-captioning) this is pytorch version of [this](https://huggingface.co/ydshieh/vit-gpt2-coco-en-ckpts).
# The Illustrated Image Captioning using transformers
* https://ankur3107.github.io/blogs/the-illustrated-image-captioning-using-transformers/
# Sample running code
```python
from transformers import VisionEncoderDecoderModel, ViTImageProcessor, AutoTokenizer
import torch
from PIL import Image
model = VisionEncoderDecoderModel.from_pretrained("nlpconnect/vit-gpt2-image-captioning")
feature_extractor = ViTImageProcessor.from_pretrained("nlpconnect/vit-gpt2-image-captioning")
tokenizer = AutoTokenizer.from_pretrained("nlpconnect/vit-gpt2-image-captioning")
device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
model.to(device)
max_length = 16
num_beams = 4
gen_kwargs = {"max_length": max_length, "num_beams": num_beams}
def predict_step(image_paths):
images = []
for image_path in image_paths:
i_image = Image.open(image_path)
if i_image.mode != "RGB":
i_image = i_image.convert(mode="RGB")
images.append(i_image)
pixel_values = feature_extractor(images=images, return_tensors="pt").pixel_values
pixel_values = pixel_values.to(device)
output_ids = model.generate(pixel_values, **gen_kwargs)
preds = tokenizer.batch_decode(output_ids, skip_special_tokens=True)
preds = [pred.strip() for pred in preds]
return preds
predict_step(['doctor.e16ba4e4.jpg']) # ['a woman in a hospital bed with a woman in a hospital bed']
```
# Sample running code using transformers pipeline
```python
from transformers import pipeline
image_to_text = pipeline("image-to-text", model="nlpconnect/vit-gpt2-image-captioning")
image_to_text("https://ankur3107.github.io/assets/images/image-captioning-example.png")
# [{'generated_text': 'a soccer game with a player jumping to catch the ball '}]
```
# Contact for any help
* https://huggingface.co/ankur310794
* https://twitter.com/ankur310794
* http://github.com/ankur3107
* https://www.linkedin.com/in/ankur310794 | null |
Salesforce/instructblip-vicuna-7b |
# InstructBLIP model
InstructBLIP model using [Vicuna-7b](https://github.com/lm-sys/FastChat#model-weights) as language model. InstructBLIP was introduced in the paper [InstructBLIP: Towards General-purpose Vision-Language Models with Instruction Tuning](https://arxiv.org/abs/2305.06500) by Dai et al.
Disclaimer: The team releasing InstructBLIP did not write a model card for this model so this model card has been written by the Hugging Face team.
## Model description
InstructBLIP is a visual instruction tuned version of [BLIP-2](https://huggingface.co/docs/transformers/main/model_doc/blip-2). Refer to the paper for details.
## Intended uses & limitations
Usage is as follows:
```
from transformers import InstructBlipProcessor, InstructBlipForConditionalGeneration
import torch
from PIL import Image
import requests
model = InstructBlipForConditionalGeneration.from_pretrained("Salesforce/instructblip-vicuna-7b")
processor = InstructBlipProcessor.from_pretrained("Salesforce/instructblip-vicuna-7b")
device = "cuda" if torch.cuda.is_available() else "cpu"
model.to(device)
url = "https://raw.githubusercontent.com/salesforce/LAVIS/main/docs/_static/Confusing-Pictures.jpg"
image = Image.open(requests.get(url, stream=True).raw).convert("RGB")
prompt = "What is unusual about this image?"
inputs = processor(images=image, text=prompt, return_tensors="pt").to(device)
outputs = model.generate(
**inputs,
do_sample=False,
num_beams=5,
max_length=256,
min_length=1,
top_p=0.9,
repetition_penalty=1.5,
length_penalty=1.0,
temperature=1,
)
generated_text = processor.batch_decode(outputs, skip_special_tokens=True)[0].strip()
print(generated_text)
```
## Ethical Considerations
This release is for research purposes only in support of an academic paper. Our models, datasets, and code are not specifically designed or evaluated for all downstream purposes. We strongly recommend users evaluate and address potential concerns related to accuracy, safety, and fairness before deploying this model. We encourage users to consider the common limitations of AI, comply with applicable laws, and leverage best practices when selecting use cases, particularly for high-risk scenarios where errors or misuse could significantly impact people’s lives, rights, or safety. For further guidance on use cases, refer to our AUP and AI AUP.
### How to use
For code examples, we refer to the [documentation](https://huggingface.co/docs/transformers/main/en/model_doc/instructblip). | null |
ogulcanakca/blip-itu-turkish-captions-finetuned |
# Türkçe Görüntü Altyazılama: BLIP ile Bir Başlangıç Noktası
## Projeye Genel Bakış ve Katkısı
Bu proje, `Salesforce/blip-image-captioning-base` modelinin, `ituperceptron/image-captioning-turkish` veri kümesinin "long_captions" bölümünden alınan bir alt küme üzerinde **Türkçe görüntü altyazıları üretmek** amacıyla ince ayar (fine-tuning) yapılmasına odaklanmaktadır. Amaç, görseller için otomatik, akıcı ve anlamlı Türkçe açıklamalar üretebilen bir model geliştirmek ve bu alanda bir "orta karar ama değerli bir başlangıç noktası" oluşturmaktır.
Bu çalışma, özellikle Türkçe gibi, İngilizceye kıyasla daha az kaynak ve önceden eğitilmiş modele sahip dillerde VLM (Görsel Dil Modeli) yeteneklerini keşfetmeyi ve geliştirmeyi hedefler.
## Veri Kümesi
* **Adı:** Türkçe Görüntü Altyazılama Veri Kümesi
* **Kaynak:** `ituperceptron/image-captioning-turkish`
* **Yapılandırma/Split:** `long_captions`
* **İçerik:** Çeşitli görseller ve bu görsellere ait insanlar tarafından yazılmış uzun, açıklayıcı Türkçe altyazılar.
* **Kullanılan Alt Küme:** Bu projede, eğitim ve değerlendirme için 60.000 örnekten oluşan bir alt küme kullanılmıştır.
## Model
* **Temel Model:** `Salesforce/blip-image-captioning-base`
* **Görev:** Görüntüden Metne Çeviri (Image-to-Text) / Görüntü Altyazılama (Image Captioning)
BLIP (Bootstrapping Language-Image Pre-training), görüntü ve metin arasındaki ilişkiyi anlamak ve bu ilişkiden yola çıkarak metin üretmek üzere tasarlanmış güçlü bir VLM mimarisidir.
## Temel Özellikler ve Yaklaşım
* `Salesforce/blip-image-captioning-base` modeli, Türkçe altyazılama görevine adapte etmek için `ituperceptron/image-captioning-turkish` veri kümesi üzerinde fine tunining yapılmıştır.
* Eğitimde, P100 GPU gibi kısıtlı kaynaklarda çalışabilmek için bellek optimizasyon teknikleri (örn: `per_device_eval_batch_size=1`, FP16 eğitimi, `PYTORCH_CUDA_ALLOC_CONF` ortam değişkeni, üretilen maksimum token sayısının sınırlandırılması) kullanılmıştır.
* Hugging Face `Transformers`, `Datasets` ve `Evaluate` kütüphaneleri, veri işleme, model eğitimi ve metriklerin hesaplanması için temel araçlar olarak kullanılmıştır.
* Modelin performansı SacreBLEU, ROUGE ve METEOR gibi standart görüntü altyazılama metrikleri ile değerlendirilmiştir.
## Performans (İnce Ayar Sonrası)
Model, `ituperceptron/image-captioning-turkish` veri kümesinin **test seti** üzerinde aşağıdaki performansı göstermiştir:
* **SacreBLEU:** `0.1129`
* **ROUGE-L:** `0.2852`
* **ROUGE-1:** `0.3486`
* **METEOR:** `0.2978`
Bu metrikler, modelin Türkçe altyazı üretme yeteneğini önemli ölçüde kazandığını ve zero-shot performansına kıyasla büyük bir gelişme kaydettiğini göstermektedir. Özellikle kısıtlı veri ve donanım kaynakları göz önüne alındığında, bu sonuçlar Türkçe VLM alanında umut verici bir başlangıç noktası sunmaktadır.
## Kullanım (Örnek)
Bu fine tuning yapılmış model, Hugging Face Hub'dan doğrudan yüklenebilir ve Türkçe görüntü altyazıları üretmek için kullanılabilir:
```python
from transformers import BlipProcessor, BlipForConditionalGeneration
from PIL import Image
import torch
# 1. Modelinizi ve İşlemcinizi Hugging Face Hub'dan Yükleyin
MODEL_ID = "ogulcanakca/blip-base-turkish-image-captioning" # modelimizin Hub ID'si
try:
processor = BlipProcessor.from_pretrained(MODEL_ID)
model = BlipForConditionalGeneration.from_pretrained(MODEL_ID)
print(f"'{MODEL_ID}' başarıyla yüklendi.")
except Exception as e:
print(f"Hata: Model veya Processor yüklenemedi: {e}")
raise
model.eval()
device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
model.to(device)
print(f"Model '{device}' cihazına taşındı ve değerlendirme modunda.")
# 2. Altyazı Üretmek İstediğiniz Görüntüyü Yükleyin
image_path = "path/to/your/image.jpg" # <<< KENDİ GÖRÜNTÜNÜZÜN YOLUNU GİRİN
try:
image_pil = Image.open(image_path).convert("RGB")
print(f"'{image_path}' başarıyla yüklendi.")
except FileNotFoundError:
print(f"HATA: Görüntü '{image_path}' bulunamadı.")
exit()
# 3. Görüntüyü İşleyin ve Altyazı Üretin
# BLIP altyazılama modelleri genellikle text prompt'a ihtiyaç duymaz.
inputs = processor(images=image_pil, return_tensors="pt").to(device, model.dtype) # model.dtype ile aynı hassasiyette
print("\nModel ile altyazı üretiliyor...")
with torch.no_grad():
generated_ids = model.generate(
**inputs,
max_length=70, # Eğitimdeki TEXT_MAX_LENGTH ile uyumlu veya istediğiniz bir değer
# num_beams=3, # İsteğe bağlı: Daha kaliteli ama yavaş üretim için
# early_stopping=True # num_beams > 1 ise
)
generated_caption = processor.decode(generated_ids[0], skip_special_tokens=True)
print("\nÜretilen Türkçe Altyazı:")
print(generated_caption)
# Görüntüyü göstermek için (isteğe bağlı):
# import matplotlib.pyplot as plt
# plt.imshow(image_pil)
# plt.title(generated_caption)
# plt.axis('off')
# plt.show()
```
## Eğitim Hiperparametreleri (Örnek)
* Base Model: Salesforce/blip-image-captioning-base
* Learning Rate: 2e-5
* Training Epochs: 3
* Training Batch Size per Device: 4
* Evaluation Batch Size per Device: 1
* Gradient Accumulation Steps: 4 (Effective batch size: 16)
* Warmup Ratio: 0.1
* Weight Decay: 0.01
* Optimizer: AdamW (adamw_torch)
* Mixed Precision: FP16
* Maximum Text Length (Training): 70
* This approach was taken in anticipation that it would become disconnected from context as the length increased.
## Ortam Bilgileri ve Süreler
* Model Boyutu: Salesforce/blip-image-captioning-base (~990MB pytorch_model.bin olarak)
* Veri Kümesi: ituperceptron/image-captioning-turkish (long_captions split'inden 60.000 örnek)
* GPU: NVIDIA P100 (Kaggle üzerinde)
* Toplam Eğitim Süresi (3 epoch için): 6:57:27.93
* Saniyede eğitilen örnek sayısı: 7.186
* Saniyede atılan eğitim adımı: 0.449
* Çıkarım Hızı: Yaklaşık 5 örnek/saniye (Trainer.predict() ile test seti üzerinde (NVIDIA P100))
### CITATION
```
@misc{ogulcanakca_blip_turkish_captioning_2025,
author = {Oğulcan Akca},
title = {Fine-tuned BLIP-Base for Turkish Image Captioning},
year = {2025},
publisher = {Hugging Face},
journal = {Hugging Face Model Hub},
howpublished = {{[https://huggingface.co/ogulcanakca/blip-base-turkish-image-captioning](https://huggingface.co/ogulcanakca/blip-base-turkish-image-captioning)}}
}
```
* BLIP:
```
@inproceedings{li2022blip,
title={BLIP: Bootstrapping Language-Image Pre-training for Unified Vision-Language Understanding and Generation},
author={Li, Junnan and Li, Dongxu and Xiong, Caiming and Hoi, Steven},
booktitle={International Conference on Machine Learning (ICML)},
year={2022}
}
```
## Notebook'lar
* [blip-turkish-image-captioning-0shot noteobok'u](https://www.kaggle.com/code/oulcanakca/blip-turkish-image-captioning-0shot)
* [blip-turkish-image-captioning-fine-tune notebook'u](https://www.kaggle.com/code/oulcanakca/blip-turkish-image-captioning-fine-tune)
* [blip-turkish-image-captioning-test notebook'u](https://www.kaggle.com/code/oulcanakca/blip-turkish-image-captioning-test)
* [blip-turkish-image-captioning-inference notebook'u](https://www.kaggle.com/code/oulcanakca/blip-turkish-image-captioning-inference)
| null |
adalbertojunior/image_captioning_portuguese |
Image Captioning in Portuguese trained with ViT and GPT2
[DEMO](https://huggingface.co/spaces/adalbertojunior/image_captioning_portuguese)
Research supported with Cloud TPUs from Google's TPU Research Cloud (TRC) | null |
deepklarity/poster2plot |
# Poster2Plot
An image captioning model to generate movie/t.v show plot from poster. It generates decent plots but is no way perfect. We are still working on improving the model.
## Live demo on Hugging Face Spaces: https://huggingface.co/spaces/deepklarity/poster2plot
# Model Details
The base model uses a Vision Transformer (ViT) model as an image encoder and GPT-2 as a decoder.
We used the following models:
* Encoder: [google/vit-base-patch16-224-in21k](https://huggingface.co/google/vit-base-patch16-224-in21k)
* Decoder: [gpt2](https://huggingface.co/gpt2)
# Datasets
Publicly available IMDb datasets were used to train the model.
# How to use
## In PyTorch
```python
import torch
import re
import requests
from PIL import Image
from transformers import AutoTokenizer, AutoFeatureExtractor, VisionEncoderDecoderModel
# Pattern to ignore all the text after 2 or more full stops
regex_pattern = "[.]{2,}"
def post_process(text):
try:
text = text.strip()
text = re.split(regex_pattern, text)[0]
except Exception as e:
print(e)
pass
return text
def predict(image, max_length=64, num_beams=4):
pixel_values = feature_extractor(images=image, return_tensors="pt").pixel_values
pixel_values = pixel_values.to(device)
with torch.no_grad():
output_ids = model.generate(
pixel_values,
max_length=max_length,
num_beams=num_beams,
return_dict_in_generate=True,
).sequences
preds = tokenizer.batch_decode(output_ids, skip_special_tokens=True)
pred = post_process(preds[0])
return pred
model_name_or_path = "deepklarity/poster2plot"
device = torch.device("cuda:0" if torch.cuda.is_available() else "cpu")
# Load model.
model = VisionEncoderDecoderModel.from_pretrained(model_name_or_path)
model.to(device)
print("Loaded model")
feature_extractor = AutoFeatureExtractor.from_pretrained(model.encoder.name_or_path)
print("Loaded feature_extractor")
tokenizer = AutoTokenizer.from_pretrained(model.decoder.name_or_path, use_fast=True)
if model.decoder.name_or_path == "gpt2":
tokenizer.pad_token = tokenizer.eos_token
print("Loaded tokenizer")
url = "https://upload.wikimedia.org/wikipedia/en/2/26/Moana_Teaser_Poster.jpg"
with Image.open(requests.get(url, stream=True).raw) as image:
pred = predict(image)
print(pred)
```
| null |
gagan3012/ViTGPT2I2A |
<!-- This model card has been generated automatically according to the information the Trainer had access to. You
should probably proofread and complete it, then remove this comment. -->
# ViTGPT2I2A
This model is a fine-tuned version of [google/vit-base-patch16-224-in21k](https://huggingface.co/google/vit-base-patch16-224-in21k) on the vizwiz dataset.
It achieves the following results on the evaluation set:
- Loss: 0.0708
## Model description
More information needed
## Intended uses & limitations
More information needed
## Training and evaluation data
More information needed
## Training procedure
### Training hyperparameters
The following hyperparameters were used during training:
- learning_rate: 2e-05
- train_batch_size: 2
- eval_batch_size: 2
- seed: 42
- distributed_type: multi-GPU
- num_devices: 2
- total_train_batch_size: 4
- total_eval_batch_size: 4
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: linear
- num_epochs: 5.0
- mixed_precision_training: Native AMP
### Training results
| Training Loss | Epoch | Step | Validation Loss |
|:-------------:|:-----:|:-----:|:---------------:|
| 0.1528 | 0.17 | 1000 | 0.0869 |
| 0.0899 | 0.34 | 2000 | 0.0817 |
| 0.084 | 0.51 | 3000 | 0.0790 |
| 0.0814 | 0.68 | 4000 | 0.0773 |
| 0.0803 | 0.85 | 5000 | 0.0757 |
| 0.077 | 1.02 | 6000 | 0.0745 |
| 0.0739 | 1.19 | 7000 | 0.0740 |
| 0.0719 | 1.37 | 8000 | 0.0737 |
| 0.0717 | 1.54 | 9000 | 0.0730 |
| 0.0731 | 1.71 | 10000 | 0.0727 |
| 0.0708 | 1.88 | 11000 | 0.0720 |
| 0.0697 | 2.05 | 12000 | 0.0717 |
| 0.0655 | 2.22 | 13000 | 0.0719 |
| 0.0653 | 2.39 | 14000 | 0.0719 |
| 0.0657 | 2.56 | 15000 | 0.0712 |
| 0.0663 | 2.73 | 16000 | 0.0710 |
| 0.0654 | 2.9 | 17000 | 0.0708 |
| 0.0645 | 3.07 | 18000 | 0.0716 |
| 0.0616 | 3.24 | 19000 | 0.0712 |
| 0.0607 | 3.41 | 20000 | 0.0712 |
| 0.0611 | 3.58 | 21000 | 0.0711 |
| 0.0615 | 3.76 | 22000 | 0.0711 |
| 0.0614 | 3.93 | 23000 | 0.0710 |
| 0.0594 | 4.1 | 24000 | 0.0716 |
| 0.0587 | 4.27 | 25000 | 0.0715 |
| 0.0574 | 4.44 | 26000 | 0.0715 |
| 0.0579 | 4.61 | 27000 | 0.0715 |
| 0.0581 | 4.78 | 28000 | 0.0715 |
| 0.0579 | 4.95 | 29000 | 0.0715 |
### Framework versions
- Transformers 4.16.2
- Pytorch 1.10.2+cu113
- Datasets 1.18.3
- Tokenizers 0.11.0
| null |
bipin/image-caption-generator |
<!-- This model card has been generated automatically according to the information the Trainer had access to. You
should probably proofread and complete it, then remove this comment. -->
# Image-caption-generator
This model is trained on [Flickr8k](https://www.kaggle.com/datasets/nunenuh/flickr8k) dataset to generate captions given an image.
It achieves the following results on the evaluation set:
- eval_loss: 0.2536
- eval_runtime: 25.369
- eval_samples_per_second: 63.818
- eval_steps_per_second: 8.002
- epoch: 4.0
- step: 3236
# Running the model using transformers library
1. Load the pre-trained model from the model hub
```python
from transformers import VisionEncoderDecoderModel, ViTFeatureExtractor, AutoTokenizer
import torch
from PIL import Image
model_name = "bipin/image-caption-generator"
# load model
model = VisionEncoderDecoderModel.from_pretrained(model_name)
feature_extractor = ViTFeatureExtractor.from_pretrained(model_name)
tokenizer = AutoTokenizer.from_pretrained("gpt2")
device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
model.to(device)
```
2. Load the image for which the caption is to be generated(note: replace the value of `img_name` with image of your choice)
```python
### replace the value with your image
img_name = "flickr_data.jpg"
img = Image.open(img_name)
if img.mode != 'RGB':
img = img.convert(mode="RGB")
```
3. Pre-process the image
```python
pixel_values = feature_extractor(images=[img], return_tensors="pt").pixel_values
pixel_values = pixel_values.to(device)
```
4. Generate the caption
```python
max_length = 128
num_beams = 4
# get model prediction
output_ids = model.generate(pixel_values, num_beams=num_beams, max_length=max_length)
# decode the generated prediction
preds = tokenizer.decode(output_ids[0], skip_special_tokens=True)
print(preds)
```
## Training procedure
The procedure used to train this model can be found [here](https://bipinkrishnan.github.io/ml-recipe-book/image_captioning.html).
### Training hyperparameters
The following hyperparameters were used during training:
- learning_rate: 5e-05
- train_batch_size: 8
- eval_batch_size: 8
- seed: 42
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: linear
- num_epochs: 5
### Framework versions
- Transformers 4.16.2
- Pytorch 1.9.1
- Datasets 1.18.4
- Tokenizers 0.11.6
| null |
yuewu/toc_titler | A model that inputs chemistry journal article table of contents (ToC) images and generates appropriate titles. Trained on all JACS ToCs and titles. | null |
dhansmair/flamingo-tiny | Flamingo Model (tiny version) pretrained on Image Captioning on the Conceptual Captions (3M) dataset.
Source Code: https://github.com/dhansmair/flamingo-mini
Demo Space: https://huggingface.co/spaces/dhansmair/flamingo-tiny-cap
Flamingo-mini: https://huggingface.co/spaces/dhansmair/flamingo-mini-cap
| null |
dhansmair/flamingo-mini | Flamingo Model pretrained on Image Captioning on the Conceptual Captions (3M) dataset.
Source Code: https://github.com/dhansmair/flamingo-mini
Demo Space: https://huggingface.co/spaces/dhansmair/flamingo-mini-cap
Flamingo-tiny: https://huggingface.co/spaces/dhansmair/flamingo-tiny-cap
| null |
Zayn/AICVTG_What_if_a_machine_could_create_captions_automatically |
This is an image captioning model training by Zayn
```python
from transformers import VisionEncoderDecoderModel, ViTFeatureExtractor, AutoTokenizer
model = VisionEncoderDecoderModel.from_pretrained("Zayn/AICVTG_What_if_a_machine_could_create_captions_automatically")
feature_extractor = ViTFeatureExtractor.from_pretrained("Zayn/AICVTG_What_if_a_machine_could_create_captions_automatically")
tokenizer = AutoTokenizer.from_pretrained("Zayn/AICVTG_What_if_a_machine_could_create_captions_automatically")
device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
model.to(device)
max_length = 20
num_beams = 8
gen_kwargs = {"max_length": max_length, "num_beams": num_beams}
def predict_step(image_paths):
images = []
for image_path in image_paths:
i_image = Image.open(image_path)
if i_image.mode != "RGB":
i_image = i_image.convert(mode="RGB")
images.append(i_image)
pixel_values = feature_extractor(images=images, return_tensors="pt").pixel_values
pixel_values = pixel_values.to(device)
output_ids = model.generate(pixel_values, **gen_kwargs)
preds = tokenizer.batch_decode(output_ids, skip_special_tokens=True)
preds = [pred.strip() for pred in preds]
return preds
predict_step(['Image URL.jpg'])
| null |
microsoft/git-base-coco |
# GIT (GenerativeImage2Text), base-sized, fine-tuned on COCO
GIT (short for GenerativeImage2Text) model, base-sized version, fine-tuned on COCO. It was introduced in the paper [GIT: A Generative Image-to-text Transformer for Vision and Language](https://arxiv.org/abs/2205.14100) by Wang et al. and first released in [this repository](https://github.com/microsoft/GenerativeImage2Text).
Disclaimer: The team releasing GIT did not write a model card for this model so this model card has been written by the Hugging Face team.
## Model description
GIT is a Transformer decoder conditioned on both CLIP image tokens and text tokens. The model is trained using "teacher forcing" on a lot of (image, text) pairs.
The goal for the model is simply to predict the next text token, giving the image tokens and previous text tokens.
The model has full access to (i.e. a bidirectional attention mask is used for) the image patch tokens, but only has access to the previous text tokens (i.e. a causal attention mask is used for the text tokens) when predicting the next text token.
This allows the model to be used for tasks like:
- image and video captioning
- visual question answering (VQA) on images and videos
- even image classification (by simply conditioning the model on the image and asking it to generate a class for it in text).
## Intended uses & limitations
You can use the raw model for image captioning. See the [model hub](https://huggingface.co/models?search=microsoft/git) to look for
fine-tuned versions on a task that interests you.
### How to use
For code examples, we refer to the [documentation](https://huggingface.co/docs/transformers/main/model_doc/git#transformers.GitForCausalLM.forward.example).
## Training data
From the paper:
> We collect 0.8B image-text pairs for pre-training, which include COCO (Lin et al., 2014), Conceptual Captions
(CC3M) (Sharma et al., 2018), SBU (Ordonez et al., 2011), Visual Genome (VG) (Krishna et al., 2016),
Conceptual Captions (CC12M) (Changpinyo et al., 2021), ALT200M (Hu et al., 2021a), and an extra 0.6B
data following a similar collection procedure in Hu et al. (2021a).
=> however this is for the model referred to as "GIT" in the paper, which is not open-sourced.
This checkpoint is "GIT-base", which is a smaller variant of GIT trained on 10 million image-text pairs.
Next, the model was fine-tuned on COCO.
See table 11 in the [paper](https://arxiv.org/abs/2205.14100) for more details.
### Preprocessing
We refer to the original repo regarding details for preprocessing during training.
During validation, one resizes the shorter edge of each image, after which center cropping is performed to a fixed-size resolution. Next, frames are normalized across the RGB channels with the ImageNet mean and standard deviation.
## Evaluation results
For evaluation results, we refer readers to the [paper](https://arxiv.org/abs/2205.14100). | null |
microsoft/git-base-textcaps |
# GIT (GenerativeImage2Text), base-sized, fine-tuned on TextCaps
GIT (short for GenerativeImage2Text) model, base-sized version, fine-tuned on TextCaps. It was introduced in the paper [GIT: A Generative Image-to-text Transformer for Vision and Language](https://arxiv.org/abs/2205.14100) by Wang et al. and first released in [this repository](https://github.com/microsoft/GenerativeImage2Text).
Disclaimer: The team releasing GIT did not write a model card for this model so this model card has been written by the Hugging Face team.
## Model description
GIT is a Transformer decoder conditioned on both CLIP image tokens and text tokens. The model is trained using "teacher forcing" on a lot of (image, text) pairs.
The goal for the model is simply to predict the next text token, giving the image tokens and previous text tokens.
The model has full access to (i.e. a bidirectional attention mask is used for) the image patch tokens, but only has access to the previous text tokens (i.e. a causal attention mask is used for the text tokens) when predicting the next text token.
This allows the model to be used for tasks like:
- image and video captioning
- visual question answering (VQA) on images and videos
- even image classification (by simply conditioning the model on the image and asking it to generate a class for it in text).
## Intended uses & limitations
You can use the raw model for image captioning. See the [model hub](https://huggingface.co/models?search=microsoft/git) to look for
fine-tuned versions on a task that interests you.
### How to use
For code examples, we refer to the [documentation](https://huggingface.co/transformers/main/model_doc/git.html).
## Training data
From the paper:
> We collect 0.8B image-text pairs for pre-training, which include COCO (Lin et al., 2014), Conceptual Captions
(CC3M) (Sharma et al., 2018), SBU (Ordonez et al., 2011), Visual Genome (VG) (Krishna et al., 2016),
Conceptual Captions (CC12M) (Changpinyo et al., 2021), ALT200M (Hu et al., 2021a), and an extra 0.6B
data following a similar collection procedure in Hu et al. (2021a).
=> however this is for the model referred to as "GIT" in the paper, which is not open-sourced.
This checkpoint is "GIT-base", which is a smaller variant of GIT trained on 10 million image-text pairs.
Next, the model was fine-tuned on TextCaps.
See table 11 in the [paper](https://arxiv.org/abs/2205.14100) for more details.
### Preprocessing
We refer to the original repo regarding details for preprocessing during training.
During validation, one resizes the shorter edge of each image, after which center cropping is performed to a fixed-size resolution. Next, frames are normalized across the RGB channels with the ImageNet mean and standard deviation.
## Evaluation results
For evaluation results, we refer readers to the [paper](https://arxiv.org/abs/2205.14100). | null |
microsoft/git-large |
# GIT (GenerativeImage2Text), large-sized
GIT (short for GenerativeImage2Text) model, large-sized version. It was introduced in the paper [GIT: A Generative Image-to-text Transformer for Vision and Language](https://arxiv.org/abs/2205.14100) by Wang et al. and first released in [this repository](https://github.com/microsoft/GenerativeImage2Text).
Disclaimer: The team releasing GIT did not write a model card for this model so this model card has been written by the Hugging Face team.
## Model description
GIT is a Transformer decoder conditioned on both CLIP image tokens and text tokens. The model is trained using "teacher forcing" on a lot of (image, text) pairs.
The goal for the model is simply to predict the next text token, giving the image tokens and previous text tokens.
The model has full access to (i.e. a bidirectional attention mask is used for) the image patch tokens, but only has access to the previous text tokens (i.e. a causal attention mask is used for the text tokens) when predicting the next text token.
This allows the model to be used for tasks like:
- image and video captioning
- visual question answering (VQA) on images and videos
- even image classification (by simply conditioning the model on the image and asking it to generate a class for it in text).
## Intended uses & limitations
You can use the raw model for image captioning. See the [model hub](https://huggingface.co/models?search=microsoft/git) to look for
fine-tuned versions on a task that interests you.
### How to use
For code examples, we refer to the [documentation](https://huggingface.co/transformers/main/model_doc/git.html).
## Training data
From the paper:
> We collect 0.8B image-text pairs for pre-training, which include COCO (Lin et al., 2014), Conceptual Captions
(CC3M) (Sharma et al., 2018), SBU (Ordonez et al., 2011), Visual Genome (VG) (Krishna et al., 2016),
Conceptual Captions (CC12M) (Changpinyo et al., 2021), ALT200M (Hu et al., 2021a), and an extra 0.6B
data following a similar collection procedure in Hu et al. (2021a).
=> however this is for the model referred to as "GIT" in the paper, which is not open-sourced.
This checkpoint is "GIT-large", which is a smaller variant of GIT trained on 20 million image-text pairs.
See table 11 in the [paper](https://arxiv.org/abs/2205.14100) for more details.
### Preprocessing
We refer to the original repo regarding details for preprocessing during training.
During validation, one resizes the shorter edge of each image, after which center cropping is performed to a fixed-size resolution. Next, frames are normalized across the RGB channels with the ImageNet mean and standard deviation.
## Evaluation results
For evaluation results, we refer readers to the [paper](https://arxiv.org/abs/2205.14100). | null |
microsoft/git-large-coco |
# GIT (GenerativeImage2Text), large-sized, fine-tuned on COCO
GIT (short for GenerativeImage2Text) model, large-sized version, fine-tuned on COCO. It was introduced in the paper [GIT: A Generative Image-to-text Transformer for Vision and Language](https://arxiv.org/abs/2205.14100) by Wang et al. and first released in [this repository](https://github.com/microsoft/GenerativeImage2Text).
Disclaimer: The team releasing GIT did not write a model card for this model so this model card has been written by the Hugging Face team.
## Model description
GIT is a Transformer decoder conditioned on both CLIP image tokens and text tokens. The model is trained using "teacher forcing" on a lot of (image, text) pairs.
The goal for the model is simply to predict the next text token, giving the image tokens and previous text tokens.
The model has full access to (i.e. a bidirectional attention mask is used for) the image patch tokens, but only has access to the previous text tokens (i.e. a causal attention mask is used for the text tokens) when predicting the next text token.
This allows the model to be used for tasks like:
- image and video captioning
- visual question answering (VQA) on images and videos
- even image classification (by simply conditioning the model on the image and asking it to generate a class for it in text).
## Intended uses & limitations
You can use the raw model for image captioning. See the [model hub](https://huggingface.co/models?search=microsoft/git) to look for
fine-tuned versions on a task that interests you.
### How to use
For code examples, we refer to the [documentation](https://huggingface.co/docs/transformers/main/model_doc/git#transformers.GitForCausalLM.forward.example).
## Training data
From the paper:
> We collect 0.8B image-text pairs for pre-training, which include COCO (Lin et al., 2014), Conceptual Captions
(CC3M) (Sharma et al., 2018), SBU (Ordonez et al., 2011), Visual Genome (VG) (Krishna et al., 2016),
Conceptual Captions (CC12M) (Changpinyo et al., 2021), ALT200M (Hu et al., 2021a), and an extra 0.6B
data following a similar collection procedure in Hu et al. (2021a).
=> however this is for the model referred to as "GIT" in the paper, which is not open-sourced.
This checkpoint is "GIT-large", which is a smaller variant of GIT trained on 20 million image-text pairs.
Next, the model was fine-tuned on COCO.
See table 11 in the [paper](https://arxiv.org/abs/2205.14100) for more details.
### Preprocessing
We refer to the original repo regarding details for preprocessing during training.
During validation, one resizes the shorter edge of each image, after which center cropping is performed to a fixed-size resolution. Next, frames are normalized across the RGB channels with the ImageNet mean and standard deviation.
## Evaluation results
For evaluation results, we refer readers to the [paper](https://arxiv.org/abs/2205.14100). | null |
microsoft/git-large-textcaps |
# GIT (GenerativeImage2Text), large-sized, fine-tuned on TextCaps
GIT (short for GenerativeImage2Text) model, large-sized version, fine-tuned on TextCaps. It was introduced in the paper [GIT: A Generative Image-to-text Transformer for Vision and Language](https://arxiv.org/abs/2205.14100) by Wang et al. and first released in [this repository](https://github.com/microsoft/GenerativeImage2Text).
Disclaimer: The team releasing GIT did not write a model card for this model so this model card has been written by the Hugging Face team.
## Model description
GIT is a Transformer decoder conditioned on both CLIP image tokens and text tokens. The model is trained using "teacher forcing" on a lot of (image, text) pairs.
The goal for the model is simply to predict the next text token, giving the image tokens and previous text tokens.
The model has full access to (i.e. a bidirectional attention mask is used for) the image patch tokens, but only has access to the previous text tokens (i.e. a causal attention mask is used for the text tokens) when predicting the next text token.
This allows the model to be used for tasks like:
- image and video captioning
- visual question answering (VQA) on images and videos
- even image classification (by simply conditioning the model on the image and asking it to generate a class for it in text).
## Intended uses & limitations
You can use the raw model for image captioning. See the [model hub](https://huggingface.co/models?search=microsoft/git) to look for
fine-tuned versions on a task that interests you.
### How to use
For code examples, we refer to the [documentation](https://huggingface.co/transformers/main/model_doc/git.html).
## Training data
From the paper:
> We collect 0.8B image-text pairs for pre-training, which include COCO (Lin et al., 2014), Conceptual Captions
(CC3M) (Sharma et al., 2018), SBU (Ordonez et al., 2011), Visual Genome (VG) (Krishna et al., 2016),
Conceptual Captions (CC12M) (Changpinyo et al., 2021), ALT200M (Hu et al., 2021a), and an extra 0.6B
data following a similar collection procedure in Hu et al. (2021a).
=> however this is for the model referred to as "GIT" in the paper, which is not open-sourced.
This checkpoint is "GIT-large", which is a smaller variant of GIT trained on 20 million image-text pairs.
Next, the model was fine-tuned on TextCaps.
See table 11 in the [paper](https://arxiv.org/abs/2205.14100) for more details.
### Preprocessing
We refer to the original repo regarding details for preprocessing during training.
During validation, one resizes the shorter edge of each image, after which center cropping is performed to a fixed-size resolution. Next, frames are normalized across the RGB channels with the ImageNet mean and standard deviation.
## Evaluation results
For evaluation results, we refer readers to the [paper](https://arxiv.org/abs/2205.14100). | null |
ybelkada/blip-image-captioning-base-football-finetuned |
# BLIP: Bootstrapping Language-Image Pre-training for Unified Vision-Language Understanding and Generation
Model card for image captioning pretrained on COCO dataset - base architecture (with ViT base backbone) - and fine-tuned on
[football dataset](https://huggingface.co/datasets/ybelkada/football-dataset).
Google Colab notebook for fine-tuning: https://colab.research.google.com/drive/1lbqiSiA0sDF7JDWPeS0tccrM85LloVha?usp=sharing
|  |
|:--:|
| <b> Pull figure from BLIP official repo | Image source: https://github.com/salesforce/BLIP </b>|
## TL;DR
Authors from the [paper](https://arxiv.org/abs/2201.12086) write in the abstract:
*Vision-Language Pre-training (VLP) has advanced the performance for many vision-language tasks. However, most existing pre-trained models only excel in either understanding-based tasks or generation-based tasks. Furthermore, performance improvement has been largely achieved by scaling up the dataset with noisy image-text pairs collected from the web, which is a suboptimal source of supervision. In this paper, we propose BLIP, a new VLP framework which transfers flexibly to both vision-language understanding and generation tasks. BLIP effectively utilizes the noisy web data by bootstrapping the captions, where a captioner generates synthetic captions and a filter removes the noisy ones. We achieve state-of-the-art results on a wide range of vision-language tasks, such as image-text retrieval (+2.7% in average recall@1), image captioning (+2.8% in CIDEr), and VQA (+1.6% in VQA score). BLIP also demonstrates strong generalization ability when directly transferred to videolanguage tasks in a zero-shot manner. Code, models, and datasets are released.*
## Usage
You can use this model for conditional and un-conditional image captioning
### Using the Pytorch model
#### Running the model on CPU
<details>
<summary> Click to expand </summary>
```python
import requests
from PIL import Image
from transformers import BlipProcessor, BlipForConditionalGeneration
processor = BlipProcessor.from_pretrained("ybelkada/blip-image-captioning-base")
model = BlipForConditionalGeneration.from_pretrained("ybelkada/blip-image-captioning-base")
img_url = 'https://storage.googleapis.com/sfr-vision-language-research/BLIP/demo.jpg'
raw_image = Image.open(requests.get(img_url, stream=True).raw).convert('RGB')
# conditional image captioning
text = "a photography of"
inputs = processor(raw_image, text, return_tensors="pt")
out = model.generate(**inputs)
print(processor.decode(out[0], skip_special_tokens=True))
# >>> a photography of a woman and her dog
# unconditional image captioning
inputs = processor(raw_image, return_tensors="pt")
out = model.generate(**inputs)
print(processor.decode(out[0], skip_special_tokens=True))
>>> a woman sitting on the beach with her dog
```
</details>
#### Running the model on GPU
##### In full precision
<details>
<summary> Click to expand </summary>
```python
import requests
from PIL import Image
from transformers import BlipProcessor, BlipForConditionalGeneration
processor = BlipProcessor.from_pretrained("Salesforce/blip-image-captioning-base")
model = BlipForConditionalGeneration.from_pretrained("Salesfoce/blip-image-captioning-base").to("cuda")
img_url = 'https://storage.googleapis.com/sfr-vision-language-research/BLIP/demo.jpg'
raw_image = Image.open(requests.get(img_url, stream=True).raw).convert('RGB')
# conditional image captioning
text = "a photography of"
inputs = processor(raw_image, text, return_tensors="pt").to("cuda")
out = model.generate(**inputs)
print(processor.decode(out[0], skip_special_tokens=True))
# >>> a photography of a woman and her dog
# unconditional image captioning
inputs = processor(raw_image, return_tensors="pt").to("cuda")
out = model.generate(**inputs)
print(processor.decode(out[0], skip_special_tokens=True))
>>> a woman sitting on the beach with her dog
```
</details>
##### In half precision (`float16`)
<details>
<summary> Click to expand </summary>
```python
import torch
import requests
from PIL import Image
from transformers import BlipProcessor, BlipForConditionalGeneration
processor = BlipProcessor.from_pretrained("Salesforce/blip-image-captioning-base")
model = BlipForConditionalGeneration.from_pretrained("Salesforce/blip-image-captioning-base", torch_dtype=torch.float16).to("cuda")
img_url = 'https://storage.googleapis.com/sfr-vision-language-research/BLIP/demo.jpg'
raw_image = Image.open(requests.get(img_url, stream=True).raw).convert('RGB')
# conditional image captioning
text = "a photography of"
inputs = processor(raw_image, text, return_tensors="pt").to("cuda", torch.float16)
out = model.generate(**inputs)
print(processor.decode(out[0], skip_special_tokens=True))
# >>> a photography of a woman and her dog
# unconditional image captioning
inputs = processor(raw_image, return_tensors="pt").to("cuda", torch.float16)
out = model.generate(**inputs)
print(processor.decode(out[0], skip_special_tokens=True))
>>> a woman sitting on the beach with her dog
```
</details>
## BibTex and citation info
```
@misc{https://doi.org/10.48550/arxiv.2201.12086,
doi = {10.48550/ARXIV.2201.12086},
url = {https://arxiv.org/abs/2201.12086},
author = {Li, Junnan and Li, Dongxu and Xiong, Caiming and Hoi, Steven},
keywords = {Computer Vision and Pattern Recognition (cs.CV), FOS: Computer and information sciences, FOS: Computer and information sciences},
title = {BLIP: Bootstrapping Language-Image Pre-training for Unified Vision-Language Understanding and Generation},
publisher = {arXiv},
year = {2022},
copyright = {Creative Commons Attribution 4.0 International}
}
``` | null |
tuman/vit-rugpt2-image-captioning |
# First image captioning model for russian language vit-rugpt2-image-captioning
This is an image captioning model trained on translated version (en-ru) of dataset COCO2014.
# Model Details
Model was initialized `google/vit-base-patch16-224-in21k` for encoder and `sberbank-ai/rugpt3large_based_on_gpt2` for decoder.
# Metrics on test data
* Bleu: 8.672
* Bleu precision 1: 30.567
* Bleu precision 2: 7.895
* Bleu precision 3: 3.261
# Sample running code
```python
from transformers import VisionEncoderDecoderModel, ViTFeatureExtractor, AutoTokenizer
import torch
from PIL import Image
model = VisionEncoderDecoderModel.from_pretrained("vit-rugpt2-image-captioning")
feature_extractor = ViTFeatureExtractor.from_pretrained("vit-rugpt2-image-captioning")
tokenizer = AutoTokenizer.from_pretrained("vit-rugpt2-image-captioning")
device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
model.to(device)
max_length = 16
num_beams = 4
gen_kwargs = {"max_length": max_length, "num_beams": num_beams}
def predict_caption(image_paths):
images = []
for image_path in image_paths:
i_image = Image.open(image_path)
if i_image.mode != "RGB":
i_image = i_image.convert(mode="RGB")
images.append(i_image)
pixel_values = feature_extractor(images=images, return_tensors="pt").pixel_values
pixel_values = pixel_values.to(device)
output_ids = model.generate(pixel_values, **gen_kwargs)
preds = tokenizer.batch_decode(output_ids, skip_special_tokens=True)
preds = [pred.strip() for pred in preds]
return preds
predict_caption(['train2014/COCO_train2014_000000295442.jpg']) # ['Самолет на взлетно-посадочной полосе аэропорта.']
```
# Sample running code using transformers pipeline
```python
from transformers import pipeline
image_to_text = pipeline("image-to-text", model="vit-rugpt2-image-captioning")
image_to_text("train2014/COCO_train2014_000000296754.jpg") # [{'generated_text': 'Человек идет по улице с зонтом.'}]
```
# Contact for any help
* https://huggingface.co/tuman
* https://github.com/tumanov-a
* https://t.me/tumanov_av | null |
microsoft/git-large-r |
# GIT (GenerativeImage2Text), large-sized, R*
*R means "re-trained by removing some offensive captions in cc12m dataset".
GIT (short for GenerativeImage2Text) model, large-sized version. It was introduced in the paper [GIT: A Generative Image-to-text Transformer for Vision and Language](https://arxiv.org/abs/2205.14100) by Wang et al. and first released in [this repository](https://github.com/microsoft/GenerativeImage2Text).
Disclaimer: The team releasing GIT did not write a model card for this model so this model card has been written by the Hugging Face team.
## Model description
GIT is a Transformer decoder conditioned on both CLIP image tokens and text tokens. The model is trained using "teacher forcing" on a lot of (image, text) pairs.
The goal for the model is simply to predict the next text token, giving the image tokens and previous text tokens.
The model has full access to (i.e. a bidirectional attention mask is used for) the image patch tokens, but only has access to the previous text tokens (i.e. a causal attention mask is used for the text tokens) when predicting the next text token.
This allows the model to be used for tasks like:
- image and video captioning
- visual question answering (VQA) on images and videos
- even image classification (by simply conditioning the model on the image and asking it to generate a class for it in text).
## Intended uses & limitations
You can use the raw model for image captioning. See the [model hub](https://huggingface.co/models?search=microsoft/git) to look for
fine-tuned versions on a task that interests you.
### How to use
For code examples, we refer to the [documentation](https://huggingface.co/transformers/main/model_doc/git.html).
## Training data
From the paper:
> We collect 0.8B image-text pairs for pre-training, which include COCO (Lin et al., 2014), Conceptual Captions
(CC3M) (Sharma et al., 2018), SBU (Ordonez et al., 2011), Visual Genome (VG) (Krishna et al., 2016),
Conceptual Captions (CC12M) (Changpinyo et al., 2021), ALT200M (Hu et al., 2021a), and an extra 0.6B
data following a similar collection procedure in Hu et al. (2021a).
=> however this is for the model referred to as "GIT" in the paper, which is not open-sourced.
This checkpoint is "GIT-large", which is a smaller variant of GIT trained on 20 million image-text pairs.
See table 11 in the [paper](https://arxiv.org/abs/2205.14100) for more details.
### Preprocessing
We refer to the original repo regarding details for preprocessing during training.
During validation, one resizes the shorter edge of each image, after which center cropping is performed to a fixed-size resolution. Next, frames are normalized across the RGB channels with the ImageNet mean and standard deviation.
## Evaluation results
For evaluation results, we refer readers to the [paper](https://arxiv.org/abs/2205.14100). | null |
microsoft/git-large-r-coco |
# GIT (GenerativeImage2Text), large-sized, fine-tuned on COCO, R*
R = re-trained by removing some offensive captions in cc12m dataset
GIT (short for GenerativeImage2Text) model, large-sized version, fine-tuned on COCO. It was introduced in the paper [GIT: A Generative Image-to-text Transformer for Vision and Language](https://arxiv.org/abs/2205.14100) by Wang et al. and first released in [this repository](https://github.com/microsoft/GenerativeImage2Text).
Disclaimer: The team releasing GIT did not write a model card for this model so this model card has been written by the Hugging Face team.
## Model description
GIT is a Transformer decoder conditioned on both CLIP image tokens and text tokens. The model is trained using "teacher forcing" on a lot of (image, text) pairs.
The goal for the model is simply to predict the next text token, giving the image tokens and previous text tokens.
The model has full access to (i.e. a bidirectional attention mask is used for) the image patch tokens, but only has access to the previous text tokens (i.e. a causal attention mask is used for the text tokens) when predicting the next text token.
This allows the model to be used for tasks like:
- image and video captioning
- visual question answering (VQA) on images and videos
- even image classification (by simply conditioning the model on the image and asking it to generate a class for it in text).
## Intended uses & limitations
You can use the raw model for image captioning. See the [model hub](https://huggingface.co/models?search=microsoft/git) to look for
fine-tuned versions on a task that interests you.
### How to use
For code examples, we refer to the [documentation](https://huggingface.co/docs/transformers/main/model_doc/git#transformers.GitForCausalLM.forward.example).
## Training data
From the paper:
> We collect 0.8B image-text pairs for pre-training, which include COCO (Lin et al., 2014), Conceptual Captions
(CC3M) (Sharma et al., 2018), SBU (Ordonez et al., 2011), Visual Genome (VG) (Krishna et al., 2016),
Conceptual Captions (CC12M) (Changpinyo et al., 2021), ALT200M (Hu et al., 2021a), and an extra 0.6B
data following a similar collection procedure in Hu et al. (2021a).
=> however this is for the model referred to as "GIT" in the paper, which is not open-sourced.
This checkpoint is "GIT-large", which is a smaller variant of GIT trained on 20 million image-text pairs.
Next, the model was fine-tuned on COCO.
See table 11 in the [paper](https://arxiv.org/abs/2205.14100) for more details.
### Preprocessing
We refer to the original repo regarding details for preprocessing during training.
During validation, one resizes the shorter edge of each image, after which center cropping is performed to a fixed-size resolution. Next, frames are normalized across the RGB channels with the ImageNet mean and standard deviation.
## Evaluation results
For evaluation results, we refer readers to the [paper](https://arxiv.org/abs/2205.14100). | null |
microsoft/git-large-r-textcaps |
# GIT (GenerativeImage2Text), large-sized, fine-tuned on TextCaps, R*
R = re-trained by removing some offensive captions in cc12m dataset
GIT (short for GenerativeImage2Text) model, large-sized version, fine-tuned on TextCaps. It was introduced in the paper [GIT: A Generative Image-to-text Transformer for Vision and Language](https://arxiv.org/abs/2205.14100) by Wang et al. and first released in [this repository](https://github.com/microsoft/GenerativeImage2Text).
Disclaimer: The team releasing GIT did not write a model card for this model so this model card has been written by the Hugging Face team.
## Model description
GIT is a Transformer decoder conditioned on both CLIP image tokens and text tokens. The model is trained using "teacher forcing" on a lot of (image, text) pairs.
The goal for the model is simply to predict the next text token, giving the image tokens and previous text tokens.
The model has full access to (i.e. a bidirectional attention mask is used for) the image patch tokens, but only has access to the previous text tokens (i.e. a causal attention mask is used for the text tokens) when predicting the next text token.
This allows the model to be used for tasks like:
- image and video captioning
- visual question answering (VQA) on images and videos
- even image classification (by simply conditioning the model on the image and asking it to generate a class for it in text).
## Intended uses & limitations
You can use the raw model for image captioning. See the [model hub](https://huggingface.co/models?search=microsoft/git) to look for
fine-tuned versions on a task that interests you.
### How to use
For code examples, we refer to the [documentation](https://huggingface.co/transformers/main/model_doc/git.html).
## Training data
From the paper:
> We collect 0.8B image-text pairs for pre-training, which include COCO (Lin et al., 2014), Conceptual Captions
(CC3M) (Sharma et al., 2018), SBU (Ordonez et al., 2011), Visual Genome (VG) (Krishna et al., 2016),
Conceptual Captions (CC12M) (Changpinyo et al., 2021), ALT200M (Hu et al., 2021a), and an extra 0.6B
data following a similar collection procedure in Hu et al. (2021a).
=> however this is for the model referred to as "GIT" in the paper, which is not open-sourced.
This checkpoint is "GIT-large", which is a smaller variant of GIT trained on 20 million image-text pairs.
Next, the model was fine-tuned on TextCaps.
See table 11 in the [paper](https://arxiv.org/abs/2205.14100) for more details.
### Preprocessing
We refer to the original repo regarding details for preprocessing during training.
During validation, one resizes the shorter edge of each image, after which center cropping is performed to a fixed-size resolution. Next, frames are normalized across the RGB channels with the ImageNet mean and standard deviation.
## Evaluation results
For evaluation results, we refer readers to the [paper](https://arxiv.org/abs/2205.14100). | null |
tifa-benchmark/promptcap-coco-vqa | This is the repo for the paper [PromptCap: Prompt-Guided Task-Aware Image Captioning](https://arxiv.org/abs/2211.09699). This paper is accepted to ICCV 2023 as [PromptCap: Prompt-Guided Image Captioning for VQA with GPT-3](https://openaccess.thecvf.com/content/ICCV2023/html/Hu_PromptCap_Prompt-Guided_Image_Captioning_for_VQA_with_GPT-3_ICCV_2023_paper.html).
We introduce PromptCap, a captioning model that can be controlled by natural language instruction. The instruction may contain a question that the user is interested in.
For example, "what is the boy putting on?". PromptCap also supports generic caption, using the question "what does the image describe?"
PromptCap can serve as a light-weight visual plug-in (much faster than BLIP-2) for LLM like GPT-3, ChatGPT, and other foundation models like Segment Anything and DINO.
It achieves SOTA performance on COCO captioning (150 CIDEr).
When paired with GPT-3, and conditioned on user question, PromptCap get SOTA performance on knowledge-based VQA tasks (60.4% on OK-VQA and 59.6% on A-OKVQA)
# QuickStart
## Installation
```
pip install promptcap
```
Two pipelines are included. One is for image captioning, and the other is for visual question answering.
## Captioning Pipeline
Please follow the prompt format, which will give the best performance.
Generate a prompt-guided caption by following:
```python
import torch
from promptcap import PromptCap
model = PromptCap("tifa-benchmark/promptcap-coco-vqa") # also support OFA checkpoints. e.g. "OFA-Sys/ofa-large"
if torch.cuda.is_available():
model.cuda()
prompt = "please describe this image according to the given question: what piece of clothing is this boy putting on?"
image = "glove_boy.jpeg"
print(model.caption(prompt, image))
```
To try generic captioning, just use "what does the image describe?"
```python
prompt = "what does the image describe?"
image = "glove_boy.jpeg"
print(model.caption(prompt, image))
```
PromptCap also support taking OCR inputs:
```python
prompt = "please describe this image according to the given question: what year was this taken?"
image = "dvds.jpg"
ocr = "yip AE Mht juor 02/14/2012"
print(model.caption(prompt, image, ocr))
```
## Visual Question Answering Pipeline
Different from typical VQA models, which are doing classification on VQAv2, PromptCap is open-domain and can be paired with arbitrary text-QA models.
Here we provide a pipeline for combining PromptCap with UnifiedQA.
```python
import torch
from promptcap import PromptCap_VQA
# QA model support all UnifiedQA variants. e.g. "allenai/unifiedqa-v2-t5-large-1251000"
vqa_model = PromptCap_VQA(promptcap_model="tifa-benchmark/promptcap-coco-vqa", qa_model="allenai/unifiedqa-t5-base")
if torch.cuda.is_available():
vqa_model.cuda()
question = "what piece of clothing is this boy putting on?"
image = "glove_boy.jpeg"
print(vqa_model.vqa(question, image))
```
Similarly, PromptCap supports OCR inputs
```python
question = "what year was this taken?"
image = "dvds.jpg"
ocr = "yip AE Mht juor 02/14/2012"
print(vqa_model.vqa(question, image, ocr=ocr))
```
Because of the flexibility of Unifiedqa, PromptCap also supports multiple-choice VQA
```python
question = "what piece of clothing is this boy putting on?"
image = "glove_boy.jpeg"
choices = ["gloves", "socks", "shoes", "coats"]
print(vqa_model.vqa_multiple_choice(question, image, choices))
```
## Bibtex
```
@article{hu2022promptcap,
title={PromptCap: Prompt-Guided Task-Aware Image Captioning},
author={Hu, Yushi and Hua, Hang and Yang, Zhengyuan and Shi, Weijia and Smith, Noah A and Luo, Jiebo},
journal={arXiv preprint arXiv:2211.09699},
year={2022}
}
``` | null |
Salesforce/blip2-flan-t5-xl |
# BLIP-2, Flan T5-xl, pre-trained only
BLIP-2 model, leveraging [Flan T5-xl](https://huggingface.co/google/flan-t5-xl) (a large language model).
It was introduced in the paper [BLIP-2: Bootstrapping Language-Image Pre-training with Frozen Image Encoders and Large Language Models](https://arxiv.org/abs/2301.12597) by Li et al. and first released in [this repository](https://github.com/salesforce/LAVIS/tree/main/projects/blip2).
Disclaimer: The team releasing BLIP-2 did not write a model card for this model so this model card has been written by the Hugging Face team.
## Model description
BLIP-2 consists of 3 models: a CLIP-like image encoder, a Querying Transformer (Q-Former) and a large language model.
The authors initialize the weights of the image encoder and large language model from pre-trained checkpoints and keep them frozen
while training the Querying Transformer, which is a BERT-like Transformer encoder that maps a set of "query tokens" to query embeddings,
which bridge the gap between the embedding space of the image encoder and the large language model.
The goal for the model is simply to predict the next text token, giving the query embeddings and the previous text.
<img src="https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/transformers/model_doc/blip2_architecture.jpg"
alt="drawing" width="600"/>
This allows the model to be used for tasks like:
- image captioning
- visual question answering (VQA)
- chat-like conversations by feeding the image and the previous conversation as prompt to the model
## Direct Use and Downstream Use
You can use the raw model for conditional text generation given an image and optional text. See the [model hub](https://huggingface.co/models?search=Salesforce/blip) to look for
fine-tuned versions on a task that interests you.
## Bias, Risks, Limitations, and Ethical Considerations
BLIP2-FlanT5 uses off-the-shelf Flan-T5 as the language model. It inherits the same risks and limitations from [Flan-T5](https://arxiv.org/pdf/2210.11416.pdf):
> Language models, including Flan-T5, can potentially be used for language generation in a harmful way, according to Rae et al. (2021). Flan-T5 should not be used directly in any application, without a prior assessment of safety and fairness concerns specific to the application.
BLIP2 is fine-tuned on image-text datasets (e.g. [LAION](https://laion.ai/blog/laion-400-open-dataset/) ) collected from the internet. As a result the model itself is potentially vulnerable to generating equivalently inappropriate content or replicating inherent biases in the underlying data.
BLIP2 has not been tested in real world applications. It should not be directly deployed in any applications. Researchers should first carefully assess the safety and fairness of the model in relation to the specific context they’re being deployed within.
## Ethical Considerations
This release is for research purposes only in support of an academic paper. Our models, datasets, and code are not specifically designed or evaluated for all downstream purposes. We strongly recommend users evaluate and address potential concerns related to accuracy, safety, and fairness before deploying this model. We encourage users to consider the common limitations of AI, comply with applicable laws, and leverage best practices when selecting use cases, particularly for high-risk scenarios where errors or misuse could significantly impact people’s lives, rights, or safety. For further guidance on use cases, refer to our AUP and AI AUP.
### How to use
For code examples, we refer to the [documentation](https://huggingface.co/docs/transformers/main/en/model_doc/blip-2#transformers.Blip2ForConditionalGeneration.forward.example).
#### Running the model on CPU
<details>
<summary> Click to expand </summary>
```python
import requests
from PIL import Image
from transformers import BlipProcessor, Blip2ForConditionalGeneration
processor = BlipProcessor.from_pretrained("Salesforce/blip2-flan-t5-xl")
model = Blip2ForConditionalGeneration.from_pretrained("Salesforce/blip2-flan-t5-xl")
img_url = 'https://storage.googleapis.com/sfr-vision-language-research/BLIP/demo.jpg'
raw_image = Image.open(requests.get(img_url, stream=True).raw).convert('RGB')
question = "how many dogs are in the picture?"
inputs = processor(raw_image, question, return_tensors="pt")
out = model.generate(**inputs)
print(processor.decode(out[0], skip_special_tokens=True))
```
</details>
#### Running the model on GPU
##### In full precision
<details>
<summary> Click to expand </summary>
```python
# pip install accelerate
import requests
from PIL import Image
from transformers import Blip2Processor, Blip2ForConditionalGeneration
processor = Blip2Processor.from_pretrained("Salesforce/blip2-flan-t5-xl")
model = Blip2ForConditionalGeneration.from_pretrained("Salesforce/blip2-flan-t5-xl", device_map="auto")
img_url = 'https://storage.googleapis.com/sfr-vision-language-research/BLIP/demo.jpg'
raw_image = Image.open(requests.get(img_url, stream=True).raw).convert('RGB')
question = "how many dogs are in the picture?"
inputs = processor(raw_image, question, return_tensors="pt").to("cuda")
out = model.generate(**inputs)
print(processor.decode(out[0], skip_special_tokens=True))
```
</details>
##### In half precision (`float16`)
<details>
<summary> Click to expand </summary>
```python
# pip install accelerate
import torch
import requests
from PIL import Image
from transformers import Blip2Processor, Blip2ForConditionalGeneration
processor = Blip2Processor.from_pretrained("Salesforce/blip2-flan-t5-xl")
model = Blip2ForConditionalGeneration.from_pretrained("Salesforce/blip2-flan-t5-xl", torch_dtype=torch.float16, device_map="auto")
img_url = 'https://storage.googleapis.com/sfr-vision-language-research/BLIP/demo.jpg'
raw_image = Image.open(requests.get(img_url, stream=True).raw).convert('RGB')
question = "how many dogs are in the picture?"
inputs = processor(raw_image, question, return_tensors="pt").to("cuda", torch.float16)
out = model.generate(**inputs)
print(processor.decode(out[0], skip_special_tokens=True))
```
</details>
##### In 8-bit precision (`int8`)
<details>
<summary> Click to expand </summary>
```python
# pip install accelerate bitsandbytes
import torch
import requests
from PIL import Image
from transformers import Blip2Processor, Blip2ForConditionalGeneration
processor = Blip2Processor.from_pretrained("Salesforce/blip2-flan-t5-xl")
model = Blip2ForConditionalGeneration.from_pretrained("Salesforce/blip2-flan-t5-xl", load_in_8bit=True, device_map="auto")
img_url = 'https://storage.googleapis.com/sfr-vision-language-research/BLIP/demo.jpg'
raw_image = Image.open(requests.get(img_url, stream=True).raw).convert('RGB')
question = "how many dogs are in the picture?"
inputs = processor(raw_image, question, return_tensors="pt").to("cuda", torch.float16)
out = model.generate(**inputs)
print(processor.decode(out[0], skip_special_tokens=True))
```
</details> | null |
Salesforce/blip2-opt-6.7b |
# BLIP-2, OPT-6.7b, pre-trained only
BLIP-2 model, leveraging [OPT-6.7b](https://huggingface.co/facebook/opt-6.7b) (a large language model with 6.7 billion parameters).
It was introduced in the paper [BLIP-2: Bootstrapping Language-Image Pre-training with Frozen Image Encoders and Large Language Models](https://arxiv.org/abs/2301.12597) by Li et al. and first released in [this repository](https://github.com/salesforce/LAVIS/tree/main/projects/blip2).
Disclaimer: The team releasing BLIP-2 did not write a model card for this model so this model card has been written by the Hugging Face team.
## Model description
BLIP-2 consists of 3 models: a CLIP-like image encoder, a Querying Transformer (Q-Former) and a large language model.
The authors initialize the weights of the image encoder and large language model from pre-trained checkpoints and keep them frozen
while training the Querying Transformer, which is a BERT-like Transformer encoder that maps a set of "query tokens" to query embeddings,
which bridge the gap between the embedding space of the image encoder and the large language model.
The goal for the model is simply to predict the next text token, giving the query embeddings and the previous text.
<img src="https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/transformers/model_doc/blip2_architecture.jpg"
alt="drawing" width="600"/>
This allows the model to be used for tasks like:
- image captioning
- visual question answering (VQA)
- chat-like conversations by feeding the image and the previous conversation as prompt to the model
## Direct Use and Downstream Use
You can use the raw model for conditional text generation given an image and optional text. See the [model hub](https://huggingface.co/models?search=Salesforce/blip) to look for
fine-tuned versions on a task that interests you.
## Bias, Risks, Limitations, and Ethical Considerations
BLIP2-OPT uses off-the-shelf OPT as the language model. It inherits the same risks and limitations as mentioned in Meta's model card.
> Like other large language models for which the diversity (or lack thereof) of training
> data induces downstream impact on the quality of our model, OPT-175B has limitations in terms
> of bias and safety. OPT-175B can also have quality issues in terms of generation diversity and
> hallucination. In general, OPT-175B is not immune from the plethora of issues that plague modern
> large language models.
>
BLIP2 is fine-tuned on image-text datasets (e.g. [LAION](https://laion.ai/blog/laion-400-open-dataset/) ) collected from the internet. As a result the model itself is potentially vulnerable to generating equivalently inappropriate content or replicating inherent biases in the underlying data.
BLIP2 has not been tested in real world applications. It should not be directly deployed in any applications. Researchers should first carefully assess the safety and fairness of the model in relation to the specific context they’re being deployed within.
## Ethical Considerations
This release is for research purposes only in support of an academic paper. Our models, datasets, and code are not specifically designed or evaluated for all downstream purposes. We strongly recommend users evaluate and address potential concerns related to accuracy, safety, and fairness before deploying this model. We encourage users to consider the common limitations of AI, comply with applicable laws, and leverage best practices when selecting use cases, particularly for high-risk scenarios where errors or misuse could significantly impact people’s lives, rights, or safety. For further guidance on use cases, refer to our AUP and AI AUP.
### How to use
For code examples, we refer to the [documentation](https://huggingface.co/docs/transformers/main/en/model_doc/blip-2#transformers.Blip2ForConditionalGeneration.forward.example). | null |
Salesforce/blip2-opt-2.7b-coco |
# BLIP-2, OPT-2.7b, fine-tuned on COCO
BLIP-2 model, leveraging [OPT-2.7b](https://huggingface.co/facebook/opt-2.7b) (a large language model with 2.7 billion parameters).
It was introduced in the paper [BLIP-2: Bootstrapping Language-Image Pre-training with Frozen Image Encoders and Large Language Models](https://arxiv.org/abs/2301.12597) by Li et al. and first released in [this repository](https://github.com/salesforce/LAVIS/tree/main/projects/blip2).
Disclaimer: The team releasing BLIP-2 did not write a model card for this model so this model card has been written by the Hugging Face team.
## Model description
BLIP-2 consists of 3 models: a CLIP-like image encoder, a Querying Transformer (Q-Former) and a large language model.
The authors initialize the weights of the image encoder and large language model from pre-trained checkpoints and keep them frozen
while training the Querying Transformer, which is a BERT-like Transformer encoder that maps a set of "query tokens" to query embeddings,
which bridge the gap between the embedding space of the image encoder and the large language model.
The goal for the model is simply to predict the next text token, giving the query embeddings and the previous text.
<img src="https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/transformers/model_doc/blip2_architecture.jpg"
alt="drawing" width="600"/>
This allows the model to be used for tasks like:
- image captioning
- visual question answering (VQA)
- chat-like conversations by feeding the image and the previous conversation as prompt to the model
## Direct Use and Downstream Use
You can use the raw model for conditional text generation given an image and optional text. See the [model hub](https://huggingface.co/models?search=Salesforce/blip) to look for
fine-tuned versions on a task that interests you.
## Bias, Risks, Limitations, and Ethical Considerations
BLIP2-OPT uses off-the-shelf OPT as the language model. It inherits the same risks and limitations as mentioned in Meta's model card.
> Like other large language models for which the diversity (or lack thereof) of training
> data induces downstream impact on the quality of our model, OPT-175B has limitations in terms
> of bias and safety. OPT-175B can also have quality issues in terms of generation diversity and
> hallucination. In general, OPT-175B is not immune from the plethora of issues that plague modern
> large language models.
>
BLIP2 is fine-tuned on image-text datasets (e.g. [LAION](https://laion.ai/blog/laion-400-open-dataset/) ) collected from the internet. As a result the model itself is potentially vulnerable to generating equivalently inappropriate content or replicating inherent biases in the underlying data.
BLIP2 has not been tested in real world applications. It should not be directly deployed in any applications. Researchers should first carefully assess the safety and fairness of the model in relation to the specific context they’re being deployed within.
## Ethical Considerations
This release is for research purposes only in support of an academic paper. Our models, datasets, and code are not specifically designed or evaluated for all downstream purposes. We strongly recommend users evaluate and address potential concerns related to accuracy, safety, and fairness before deploying this model. We encourage users to consider the common limitations of AI, comply with applicable laws, and leverage best practices when selecting use cases, particularly for high-risk scenarios where errors or misuse could significantly impact people’s lives, rights, or safety. For further guidance on use cases, refer to our AUP and AI AUP.
### How to use
For code examples, we refer to the [documentation](https://huggingface.co/docs/transformers/main/en/model_doc/blip-2#transformers.Blip2ForConditionalGeneration.forward.example). | null |
Salesforce/blip2-opt-6.7b-coco |
# BLIP-2, OPT-6.7b, fine-tuned on COCO
BLIP-2 model, leveraging [OPT-6.7b](https://huggingface.co/facebook/opt-6.7b) (a large language model with 6.7 billion parameters).
It was introduced in the paper [BLIP-2: Bootstrapping Language-Image Pre-training with Frozen Image Encoders and Large Language Models](https://arxiv.org/abs/2301.12597) by Li et al. and first released in [this repository](https://github.com/salesforce/LAVIS/tree/main/projects/blip2).
Disclaimer: The team releasing BLIP-2 did not write a model card for this model so this model card has been written by the Hugging Face team.
## Model description
BLIP-2 consists of 3 models: a CLIP-like image encoder, a Querying Transformer (Q-Former) and a large language model.
The authors initialize the weights of the image encoder and large language model from pre-trained checkpoints and keep them frozen
while training the Querying Transformer, which is a BERT-like Transformer encoder that maps a set of "query tokens" to query embeddings,
which bridge the gap between the embedding space of the image encoder and the large language model.
The goal for the model is simply to predict the next text token, giving the query embeddings and the previous text.
<img src="https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/transformers/model_doc/blip2_architecture.jpg"
alt="drawing" width="600"/>
This allows the model to be used for tasks like:
- image captioning
- visual question answering (VQA)
- chat-like conversations by feeding the image and the previous conversation as prompt to the model
## Direct Use and Downstream Use
You can use the raw model for conditional text generation given an image and optional text. See the [model hub](https://huggingface.co/models?search=Salesforce/blip) to look for
fine-tuned versions on a task that interests you.
## Bias, Risks, Limitations, and Ethical Considerations
BLIP2-OPT uses off-the-shelf OPT as the language model. It inherits the same risks and limitations as mentioned in Meta's model card.
> Like other large language models for which the diversity (or lack thereof) of training
> data induces downstream impact on the quality of our model, OPT-175B has limitations in terms
> of bias and safety. OPT-175B can also have quality issues in terms of generation diversity and
> hallucination. In general, OPT-175B is not immune from the plethora of issues that plague modern
> large language models.
>
BLIP2 is fine-tuned on image-text datasets (e.g. [LAION](https://laion.ai/blog/laion-400-open-dataset/) ) collected from the internet. As a result the model itself is potentially vulnerable to generating equivalently inappropriate content or replicating inherent biases in the underlying data.
BLIP2 has not been tested in real world applications. It should not be directly deployed in any applications. Researchers should first carefully assess the safety and fairness of the model in relation to the specific context they’re being deployed within.
## Ethical Considerations
This release is for research purposes only in support of an academic paper. Our models, datasets, and code are not specifically designed or evaluated for all downstream purposes. We strongly recommend users evaluate and address potential concerns related to accuracy, safety, and fairness before deploying this model. We encourage users to consider the common limitations of AI, comply with applicable laws, and leverage best practices when selecting use cases, particularly for high-risk scenarios where errors or misuse could significantly impact people’s lives, rights, or safety. For further guidance on use cases, refer to our AUP and AI AUP.
### How to use
For code examples, we refer to the [documentation](https://huggingface.co/docs/transformers/main/en/model_doc/blip-2#transformers.Blip2ForConditionalGeneration.forward.example). | null |
Salesforce/blip2-flan-t5-xl-coco |
# BLIP-2, Flan T5-xl, fine-tuned on COCO
BLIP-2 model, leveraging [Flan T5-xl](https://huggingface.co/google/flan-t5-xl) (a large language model).
It was introduced in the paper [BLIP-2: Bootstrapping Language-Image Pre-training with Frozen Image Encoders and Large Language Models](https://arxiv.org/abs/2301.12597) by Li et al. and first released in [this repository](https://github.com/salesforce/LAVIS/tree/main/projects/blip2).
Disclaimer: The team releasing BLIP-2 did not write a model card for this model so this model card has been written by the Hugging Face team.
## Model description
BLIP-2 consists of 3 models: a CLIP-like image encoder, a Querying Transformer (Q-Former) and a large language model.
The authors initialize the weights of the image encoder and large language model from pre-trained checkpoints and keep them frozen
while training the Querying Transformer, which is a BERT-like Transformer encoder that maps a set of "query tokens" to query embeddings,
which bridge the gap between the embedding space of the image encoder and the large language model.
The goal for the model is simply to predict the next text token, giving the query embeddings and the previous text.
<img src="https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/transformers/model_doc/blip2_architecture.jpg"
alt="drawing" width="600"/>
This allows the model to be used for tasks like:
- image captioning
- visual question answering (VQA)
- chat-like conversations by feeding the image and the previous conversation as prompt to the model
## Direct Use and Downstream Use
You can use the raw model for conditional text generation given an image and optional text. See the [model hub](https://huggingface.co/models?search=Salesforce/blip) to look for
fine-tuned versions on a task that interests you.
## Bias, Risks, Limitations, and Ethical Considerations
BLIP2-FlanT5 uses off-the-shelf Flan-T5 as the language model. It inherits the same risks and limitations from [Flan-T5](https://arxiv.org/pdf/2210.11416.pdf):
> Language models, including Flan-T5, can potentially be used for language generation in a harmful way, according to Rae et al. (2021). Flan-T5 should not be used directly in any application, without a prior assessment of safety and fairness concerns specific to the application.
BLIP2 is fine-tuned on image-text datasets (e.g. [LAION](https://laion.ai/blog/laion-400-open-dataset/) ) collected from the internet. As a result the model itself is potentially vulnerable to generating equivalently inappropriate content or replicating inherent biases in the underlying data.
BLIP2 has not been tested in real world applications. It should not be directly deployed in any applications. Researchers should first carefully assess the safety and fairness of the model in relation to the specific context they’re being deployed within.
## Ethical Considerations
This release is for research purposes only in support of an academic paper. Our models, datasets, and code are not specifically designed or evaluated for all downstream purposes. We strongly recommend users evaluate and address potential concerns related to accuracy, safety, and fairness before deploying this model. We encourage users to consider the common limitations of AI, comply with applicable laws, and leverage best practices when selecting use cases, particularly for high-risk scenarios where errors or misuse could significantly impact people’s lives, rights, or safety. For further guidance on use cases, refer to our AUP and AI AUP.
### How to use
For code examples, we refer to the [documentation](https://huggingface.co/docs/transformers/main/en/model_doc/blip-2#transformers.Blip2ForConditionalGeneration.forward.example). | null |
Salesforce/blip2-flan-t5-xxl |
# BLIP-2, Flan T5-xxl, pre-trained only
BLIP-2 model, leveraging [Flan T5-xxl](https://huggingface.co/google/flan-t5-xxl) (a large language model).
It was introduced in the paper [BLIP-2: Bootstrapping Language-Image Pre-training with Frozen Image Encoders and Large Language Models](https://arxiv.org/abs/2301.12597) by Li et al. and first released in [this repository](https://github.com/salesforce/LAVIS/tree/main/projects/blip2).
Disclaimer: The team releasing BLIP-2 did not write a model card for this model so this model card has been written by the Hugging Face team.
## Model description
BLIP-2 consists of 3 models: a CLIP-like image encoder, a Querying Transformer (Q-Former) and a large language model.
The authors initialize the weights of the image encoder and large language model from pre-trained checkpoints and keep them frozen
while training the Querying Transformer, which is a BERT-like Transformer encoder that maps a set of "query tokens" to query embeddings,
which bridge the gap between the embedding space of the image encoder and the large language model.
The goal for the model is simply to predict the next text token, giving the query embeddings and the previous text.
<img src="https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/transformers/model_doc/blip2_architecture.jpg"
alt="drawing" width="600"/>
This allows the model to be used for tasks like:
- image captioning
- visual question answering (VQA)
- chat-like conversations by feeding the image and the previous conversation as prompt to the model
## Direct Use and Downstream Use
You can use the raw model for conditional text generation given an image and optional text. See the [model hub](https://huggingface.co/models?search=Salesforce/blip) to look for
fine-tuned versions on a task that interests you.
## Bias, Risks, Limitations, and Ethical Considerations
BLIP2-FlanT5 uses off-the-shelf Flan-T5 as the language model. It inherits the same risks and limitations from [Flan-T5](https://arxiv.org/pdf/2210.11416.pdf):
> Language models, including Flan-T5, can potentially be used for language generation in a harmful way, according to Rae et al. (2021). Flan-T5 should not be used directly in any application, without a prior assessment of safety and fairness concerns specific to the application.
BLIP2 is fine-tuned on image-text datasets (e.g. [LAION](https://laion.ai/blog/laion-400-open-dataset/) ) collected from the internet. As a result the model itself is potentially vulnerable to generating equivalently inappropriate content or replicating inherent biases in the underlying data.
BLIP2 has not been tested in real world applications. It should not be directly deployed in any applications. Researchers should first carefully assess the safety and fairness of the model in relation to the specific context they’re being deployed within.
## Ethical Considerations
This release is for research purposes only in support of an academic paper. Our models, datasets, and code are not specifically designed or evaluated for all downstream purposes. We strongly recommend users evaluate and address potential concerns related to accuracy, safety, and fairness before deploying this model. We encourage users to consider the common limitations of AI, comply with applicable laws, and leverage best practices when selecting use cases, particularly for high-risk scenarios where errors or misuse could significantly impact people’s lives, rights, or safety. For further guidance on use cases, refer to our AUP and AI AUP.
### How to use
For code examples, we refer to the [documentation](https://huggingface.co/docs/transformers/main/en/model_doc/blip-2#transformers.Blip2ForConditionalGeneration.forward.example), or refer to the snippets below depending on your usecase:
#### Running the model on CPU
<details>
<summary> Click to expand </summary>
```python
import requests
from PIL import Image
from transformers import BlipProcessor, Blip2ForConditionalGeneration
processor = BlipProcessor.from_pretrained("Salesforce/blip2-flan-t5-xxl")
model = Blip2ForConditionalGeneration.from_pretrained("Salesforce/blip2-flan-t5-xxl")
img_url = 'https://storage.googleapis.com/sfr-vision-language-research/BLIP/demo.jpg'
raw_image = Image.open(requests.get(img_url, stream=True).raw).convert('RGB')
question = "how many dogs are in the picture?"
inputs = processor(raw_image, question, return_tensors="pt")
out = model.generate(**inputs)
print(processor.decode(out[0], skip_special_tokens=True))
```
</details>
#### Running the model on GPU
##### In full precision
<details>
<summary> Click to expand </summary>
```python
# pip install accelerate
import requests
from PIL import Image
from transformers import Blip2Processor, Blip2ForConditionalGeneration
processor = Blip2Processor.from_pretrained("Salesforce/blip2-flan-t5-xxl")
model = Blip2ForConditionalGeneration.from_pretrained("Salesforce/blip2-flan-t5-xxl", device_map="auto")
img_url = 'https://storage.googleapis.com/sfr-vision-language-research/BLIP/demo.jpg'
raw_image = Image.open(requests.get(img_url, stream=True).raw).convert('RGB')
question = "how many dogs are in the picture?"
inputs = processor(raw_image, question, return_tensors="pt").to("cuda")
out = model.generate(**inputs)
print(processor.decode(out[0], skip_special_tokens=True))
```
</details>
##### In half precision (`float16`)
<details>
<summary> Click to expand </summary>
```python
# pip install accelerate
import torch
import requests
from PIL import Image
from transformers import Blip2Processor, Blip2ForConditionalGeneration
processor = Blip2Processor.from_pretrained("Salesforce/blip2-flan-t5-xxl")
model = Blip2ForConditionalGeneration.from_pretrained("Salesforce/blip2-flan-t5-xxl", torch_dtype=torch.float16, device_map="auto")
img_url = 'https://storage.googleapis.com/sfr-vision-language-research/BLIP/demo.jpg'
raw_image = Image.open(requests.get(img_url, stream=True).raw).convert('RGB')
question = "how many dogs are in the picture?"
inputs = processor(raw_image, question, return_tensors="pt").to("cuda", torch.float16)
out = model.generate(**inputs)
print(processor.decode(out[0], skip_special_tokens=True))
```
</details>
##### In 8-bit precision (`int8`)
<details>
<summary> Click to expand </summary>
```python
# pip install accelerate bitsandbytes
import torch
import requests
from PIL import Image
from transformers import Blip2Processor, Blip2ForConditionalGeneration
processor = Blip2Processor.from_pretrained("Salesforce/blip2-flan-t5-xxl")
model = Blip2ForConditionalGeneration.from_pretrained("Salesforce/blip2-flan-t5-xxl", load_in_8bit=True, device_map="auto")
img_url = 'https://storage.googleapis.com/sfr-vision-language-research/BLIP/demo.jpg'
raw_image = Image.open(requests.get(img_url, stream=True).raw).convert('RGB')
question = "how many dogs are in the picture?"
inputs = processor(raw_image, question, return_tensors="pt").to("cuda", torch.float16)
out = model.generate(**inputs)
print(processor.decode(out[0], skip_special_tokens=True))
```
</details> | null |
jaimin/image_caption |
# Sample running code
```python
from transformers import VisionEncoderDecoderModel, ViTFeatureExtractor, AutoTokenizer
import torch
from PIL import Image
model = VisionEncoderDecoderModel.from_pretrained("jaimin/image_caption")
feature_extractor = ViTFeatureExtractor.from_pretrained("jaimin/image_caption")
tokenizer = AutoTokenizer.from_pretrained("jaimin/image_caption")
device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
model.to(device)
max_length = 16
num_beams = 4
gen_kwargs = {"max_length": max_length, "num_beams": num_beams}
def predict_step(image_paths):
images = []
for image_path in image_paths:
i_image = Image.open(image_path)
if i_image.mode != "RGB":
i_image = i_image.convert(mode="RGB")
images.append(i_image)
pixel_values = feature_extractor(images=images, return_tensors="pt").pixel_values
pixel_values = pixel_values.to(device)
output_ids = model.generate(pixel_values, **gen_kwargs)
preds = tokenizer.batch_decode(output_ids, skip_special_tokens=True)
preds = [pred.strip() for pred in preds]
return preds
```
# Sample running code using transformers pipeline
```python
from transformers import pipeline
image_to_text = pipeline("image-to-text", model="jaimin/image_caption")
``` | null |
Tomatolovve/DemoTest |
# nlpconnect/vit-gpt2-image-captioning
This is an image captioning model trained by @ydshieh in [flax ](https://github.com/huggingface/transformers/tree/main/examples/flax/image-captioning) this is pytorch version of [this](https://huggingface.co/ydshieh/vit-gpt2-coco-en-ckpts).
# The Illustrated Image Captioning using transformers
* https://ankur3107.github.io/blogs/the-illustrated-image-captioning-using-transformers/
# Sample running code
```python
from transformers import VisionEncoderDecoderModel, ViTImageProcessor, AutoTokenizer
import torch
from PIL import Image
model = VisionEncoderDecoderModel.from_pretrained("nlpconnect/vit-gpt2-image-captioning")
feature_extractor = ViTImageProcessor.from_pretrained("nlpconnect/vit-gpt2-image-captioning")
tokenizer = AutoTokenizer.from_pretrained("nlpconnect/vit-gpt2-image-captioning")
device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
model.to(device)
max_length = 16
num_beams = 4
gen_kwargs = {"max_length": max_length, "num_beams": num_beams}
def predict_step(image_paths):
images = []
for image_path in image_paths:
i_image = Image.open(image_path)
if i_image.mode != "RGB":
i_image = i_image.convert(mode="RGB")
images.append(i_image)
pixel_values = feature_extractor(images=images, return_tensors="pt").pixel_values
pixel_values = pixel_values.to(device)
output_ids = model.generate(pixel_values, **gen_kwargs)
preds = tokenizer.batch_decode(output_ids, skip_special_tokens=True)
preds = [pred.strip() for pred in preds]
return preds
predict_step(['doctor.e16ba4e4.jpg']) # ['a woman in a hospital bed with a woman in a hospital bed']
```
# Sample running code using transformers pipeline
```python
from transformers import pipeline
image_to_text = pipeline("image-to-text", model="nlpconnect/vit-gpt2-image-captioning")
image_to_text("https://ankur3107.github.io/assets/images/image-captioning-example.png")
# [{'generated_text': 'a soccer game with a player jumping to catch the ball '}]
```
# Contact for any help
* https://huggingface.co/ankur310794
* https://twitter.com/ankur310794
* http://github.com/ankur3107
* https://www.linkedin.com/in/ankur310794 | null |
Maciel/Muge-Image-Caption |
### 功能介绍
该模型功能主要是对图片生成文字描述。模型结构使用Encoder-Decoder结构，其中Encoder端使用BEiT模型，Decoder使用GPT模型。
使用中文Muge数据集训练语料，训练5k步，最终验证集loss为0.3737，rouge1为20.419，rouge2为7.3553，rougeL为17.3753，rougeLsum为17.376。
[Github项目地址](https://github.com/Macielyoung/Chinese-Image-Caption)
### 如何使用
```python
from transformers import VisionEncoderDecoderModel, ViTFeatureExtractor, AutoTokenizer
from PIL import Image
pretrained = "Maciel/Muge-Image-Caption"
model = VisionEncoderDecoderModel.from_pretrained(pretrained)
feature_extractor = ViTFeatureExtractor.from_pretrained(pretrained)
tokenizer = AutoTokenizer.from_pretrained(pretrained)
image_path = "https://huggingface.co/Maciel/Muge-Image-Caption/blob/main/%E9%AB%98%E8%B7%9F%E9%9E%8B.jpg"
image = Image.open(image_path)
if image.mode != "RGB":
image = image.convert("RGB")
pixel_values = feature_extractor(images=image, return_tensors="pt").pixel_values
output_ids = model.generate(pixel_values, **gen_kwargs)
preds = tokenizer.batch_decode(output_ids, skip_special_tokens=True)
preds = [pred.strip() for pred in preds]
print(preds)
```
| null |
baseplate/vit-gpt2-image-captioning |
# nlpconnect/vit-gpt2-image-captioning
This is an image captioning model trained by @ydshieh in [flax ](https://github.com/huggingface/transformers/tree/main/examples/flax/image-captioning) this is pytorch version of [this](https://huggingface.co/ydshieh/vit-gpt2-coco-en-ckpts).
# The Illustrated Image Captioning using transformers
* https://ankur3107.github.io/blogs/the-illustrated-image-captioning-using-transformers/
# Sample running code
```python
from transformers import VisionEncoderDecoderModel, ViTImageProcessor, AutoTokenizer
import torch
from PIL import Image
model = VisionEncoderDecoderModel.from_pretrained("nlpconnect/vit-gpt2-image-captioning")
feature_extractor = ViTImageProcessor.from_pretrained("nlpconnect/vit-gpt2-image-captioning")
tokenizer = AutoTokenizer.from_pretrained("nlpconnect/vit-gpt2-image-captioning")
device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
model.to(device)
max_length = 16
num_beams = 4
gen_kwargs = {"max_length": max_length, "num_beams": num_beams}
def predict_step(image_paths):
images = []
for image_path in image_paths:
i_image = Image.open(image_path)
if i_image.mode != "RGB":
i_image = i_image.convert(mode="RGB")
images.append(i_image)
pixel_values = feature_extractor(images=images, return_tensors="pt").pixel_values
pixel_values = pixel_values.to(device)
output_ids = model.generate(pixel_values, **gen_kwargs)
preds = tokenizer.batch_decode(output_ids, skip_special_tokens=True)
preds = [pred.strip() for pred in preds]
return preds
predict_step(['doctor.e16ba4e4.jpg']) # ['a woman in a hospital bed with a woman in a hospital bed']
```
# Sample running code using transformers pipeline
```python
from transformers import pipeline
image_to_text = pipeline("image-to-text", model="nlpconnect/vit-gpt2-image-captioning")
image_to_text("https://ankur3107.github.io/assets/images/image-captioning-example.png")
# [{'generated_text': 'a soccer game with a player jumping to catch the ball '}]
```
# Contact for any help
* https://huggingface.co/ankur310794
* https://twitter.com/ankur310794
* http://github.com/ankur3107
* https://www.linkedin.com/in/ankur310794 | null |
memegpt/blip2_endpoint |
# BLIP-2, OPT-2.7b, pre-trained only
BLIP-2 model, leveraging [OPT-2.7b](https://huggingface.co/facebook/opt-2.7b) (a large language model with 2.7 billion parameters).
It was introduced in the paper [BLIP-2: Bootstrapping Language-Image Pre-training with Frozen Image Encoders and Large Language Models](https://arxiv.org/abs/2301.12597) by Li et al. and first released in [this repository](https://github.com/salesforce/LAVIS/tree/main/projects/blip2).
Disclaimer: The team releasing BLIP-2 did not write a model card for this model so this model card has been written by the Hugging Face team.
## Model description
BLIP-2 consists of 3 models: a CLIP-like image encoder, a Querying Transformer (Q-Former) and a large language model.
The authors initialize the weights of the image encoder and large language model from pre-trained checkpoints and keep them frozen
while training the Querying Transformer, which is a BERT-like Transformer encoder that maps a set of "query tokens" to query embeddings,
which bridge the gap between the embedding space of the image encoder and the large language model.
The goal for the model is simply to predict the next text token, giving the query embeddings and the previous text.
<img src="https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/transformers/model_doc/blip2_architecture.jpg"
alt="drawing" width="600"/>
This allows the model to be used for tasks like:
- image captioning
- visual question answering (VQA)
- chat-like conversations by feeding the image and the previous conversation as prompt to the model
## Direct Use and Downstream Use
You can use the raw model for conditional text generation given an image and optional text. See the [model hub](https://huggingface.co/models?search=Salesforce/blip) to look for
fine-tuned versions on a task that interests you.
## Bias, Risks, Limitations, and Ethical Considerations
BLIP2-OPT uses off-the-shelf OPT as the language model. It inherits the same risks and limitations as mentioned in Meta's model card.
> Like other large language models for which the diversity (or lack thereof) of training
> data induces downstream impact on the quality of our model, OPT-175B has limitations in terms
> of bias and safety. OPT-175B can also have quality issues in terms of generation diversity and
> hallucination. In general, OPT-175B is not immune from the plethora of issues that plague modern
> large language models.
>
BLIP2 is fine-tuned on image-text datasets (e.g. [LAION](https://laion.ai/blog/laion-400-open-dataset/) ) collected from the internet. As a result the model itself is potentially vulnerable to generating equivalently inappropriate content or replicating inherent biases in the underlying data.
BLIP2 has not been tested in real world applications. It should not be directly deployed in any applications. Researchers should first carefully assess the safety and fairness of the model in relation to the specific context they’re being deployed within.
### How to use
For code examples, we refer to the [documentation](https://huggingface.co/docs/transformers/main/en/model_doc/blip-2#transformers.Blip2ForConditionalGeneration.forward.example).
#### Running the model on CPU
<details>
<summary> Click to expand </summary>
```python
import requests
from PIL import Image
from transformers import BlipProcessor, Blip2ForConditionalGeneration
processor = BlipProcessor.from_pretrained("Salesforce/blip2-opt-2.7b")
model = Blip2ForConditionalGeneration.from_pretrained("Salesforce/blip2-opt-2.7b")
img_url = 'https://storage.googleapis.com/sfr-vision-language-research/BLIP/demo.jpg'
raw_image = Image.open(requests.get(img_url, stream=True).raw).convert('RGB')
question = "how many dogs are in the picture?"
inputs = processor(raw_image, question, return_tensors="pt")
out = model.generate(**inputs)
print(processor.decode(out[0], skip_special_tokens=True))
```
</details>
#### Running the model on GPU
##### In full precision
<details>
<summary> Click to expand </summary>
```python
# pip install accelerate
import requests
from PIL import Image
from transformers import Blip2Processor, Blip2ForConditionalGeneration
processor = Blip2Processor.from_pretrained("Salesforce/blip2-opt-2.7b")
model = Blip2ForConditionalGeneration.from_pretrained("Salesforce/blip2-opt-2.7b", device_map="auto")
img_url = 'https://storage.googleapis.com/sfr-vision-language-research/BLIP/demo.jpg'
raw_image = Image.open(requests.get(img_url, stream=True).raw).convert('RGB')
question = "how many dogs are in the picture?"
inputs = processor(raw_image, question, return_tensors="pt").to("cuda")
out = model.generate(**inputs)
print(processor.decode(out[0], skip_special_tokens=True))
```
</details>
##### In half precision (`float16`)
<details>
<summary> Click to expand </summary>
```python
# pip install accelerate
import torch
import requests
from PIL import Image
from transformers import Blip2Processor, Blip2ForConditionalGeneration
processor = Blip2Processor.from_pretrained("Salesforce/blip2-opt-2.7b")
model = Blip2ForConditionalGeneration.from_pretrained("Salesforce/blip2-opt-2.7b", torch_dtype=torch.float16, device_map="auto")
img_url = 'https://storage.googleapis.com/sfr-vision-language-research/BLIP/demo.jpg'
raw_image = Image.open(requests.get(img_url, stream=True).raw).convert('RGB')
question = "how many dogs are in the picture?"
inputs = processor(raw_image, question, return_tensors="pt").to("cuda", torch.float16)
out = model.generate(**inputs)
print(processor.decode(out[0], skip_special_tokens=True))
```
</details>
##### In 8-bit precision (`int8`)
<details>
<summary> Click to expand </summary>
```python
# pip install accelerate bitsandbytes
import torch
import requests
from PIL import Image
from transformers import Blip2Processor, Blip2ForConditionalGeneration
processor = Blip2Processor.from_pretrained("Salesforce/blip2-opt-2.7b")
model = Blip2ForConditionalGeneration.from_pretrained("Salesforce/blip2-opt-2.7b", load_in_8bit=True, device_map="auto")
img_url = 'https://storage.googleapis.com/sfr-vision-language-research/BLIP/demo.jpg'
raw_image = Image.open(requests.get(img_url, stream=True).raw).convert('RGB')
question = "how many dogs are in the picture?"
inputs = processor(raw_image, question, return_tensors="pt").to("cuda", torch.float16)
out = model.generate(**inputs)
print(processor.decode(out[0], skip_special_tokens=True))
```
</details> | null |
kpyu/video-blip-opt-2.7b-ego4d |
# VideoBLIP, OPT-2.7b, fine-tuned on Ego4D
VideoBLIP model, leveraging [BLIP-2](https://arxiv.org/abs/2301.12597) with [OPT-2.7b](https://huggingface.co/facebook/opt-2.7b) (a large language model with 2.7 billion parameters) as its LLM backbone.
## Model description
VideoBLIP is an augmented BLIP-2 that can handle videos.
## Bias, Risks, Limitations, and Ethical Considerations
VideoBLIP-OPT uses off-the-shelf OPT as the language model. It inherits the same risks and limitations as mentioned in Meta's model card.
> Like other large language models for which the diversity (or lack thereof) of training
> data induces downstream impact on the quality of our model, OPT-175B has limitations in terms
> of bias and safety. OPT-175B can also have quality issues in terms of generation diversity and
> hallucination. In general, OPT-175B is not immune from the plethora of issues that plague modern
> large language models.
>
VideoBLIP has not been tested in real world applications. It should not be directly deployed in any applications. Researchers should first carefully assess the safety and fairness of the model in relation to the specific context they’re being deployed within.
### How to use
For code examples, please refer to the [official repository](https://github.com/yukw777/VideoBLIP). | null |
kpyu/video-blip-flan-t5-xl-ego4d |
# VideoBLIP, Flan T5-xl, fine-tuned on Ego4D
VideoBLIP model, leveraging [BLIP-2](https://arxiv.org/abs/2301.12597) with [Flan T5-xl](https://huggingface.co/google/flan-t5-xl) (a large language model with 2.7 billion parameters) as its LLM backbone.
## Model description
VideoBLIP is an augmented BLIP-2 that can handle videos.
## Bias, Risks, Limitations, and Ethical Considerations
VideoBLIP-OPT uses off-the-shelf Flan-T5 as the language model. It inherits the same risks and limitations from [Flan-T5](https://arxiv.org/pdf/2210.11416.pdf):
> Language models, including Flan-T5, can potentially be used for language generation in a harmful way, according to Rae et al. (2021). Flan-T5 should not be used directly in any application, without a prior assessment of safety and fairness concerns specific to the application.
VideoBLIP has not been tested in real world applications. It should not be directly deployed in any applications. Researchers should first carefully assess the safety and fairness of the model in relation to the specific context they’re being deployed within.
### How to use
For code examples, please refer to the [official repository](https://github.com/yukw777/VideoBLIP). | null |
wangjin2000/git-base-finetune |
# GIT (GenerativeImage2Text), base-sized
GIT (short for GenerativeImage2Text) model, base-sized version. It was introduced in the paper [GIT: A Generative Image-to-text Transformer for Vision and Language](https://arxiv.org/abs/2205.14100) by Wang et al. and first released in [this repository](https://github.com/microsoft/GenerativeImage2Text).
Disclaimer: The team releasing GIT did not write a model card for this model so this model card has been written by the Hugging Face team.
## Model description
GIT is a Transformer decoder conditioned on both CLIP image tokens and text tokens. The model is trained using "teacher forcing" on a lot of (image, text) pairs.
The goal for the model is simply to predict the next text token, giving the image tokens and previous text tokens.
The model has full access to (i.e. a bidirectional attention mask is used for) the image patch tokens, but only has access to the previous text tokens (i.e. a causal attention mask is used for the text tokens) when predicting the next text token.
This allows the model to be used for tasks like:
- image and video captioning
- visual question answering (VQA) on images and videos
- even image classification (by simply conditioning the model on the image and asking it to generate a class for it in text).
## Intended uses & limitations
You can use the raw model for image captioning. See the [model hub](https://huggingface.co/models?search=microsoft/git) to look for
fine-tuned versions on a task that interests you.
### How to use
For code examples, we refer to the [documentation](https://huggingface.co/docs/transformers/main/model_doc/git#transformers.GitForCausalLM.forward.example).
## Training data
From the paper:
> We collect 0.8B image-text pairs for pre-training, which include COCO (Lin et al., 2014), Conceptual Captions
(CC3M) (Sharma et al., 2018), SBU (Ordonez et al., 2011), Visual Genome (VG) (Krishna et al., 2016),
Conceptual Captions (CC12M) (Changpinyo et al., 2021), ALT200M (Hu et al., 2021a), and an extra 0.6B
data following a similar collection procedure in Hu et al. (2021a).
=> however this is for the model referred to as "GIT" in the paper, which is not open-sourced.
This checkpoint is "GIT-base", which is a smaller variant of GIT trained on 10 million image-text pairs.
See table 11 in the [paper](https://arxiv.org/abs/2205.14100) for more details.
### Preprocessing
We refer to the original repo regarding details for preprocessing during training.
During validation, one resizes the shorter edge of each image, after which center cropping is performed to a fixed-size resolution. Next, frames are normalized across the RGB channels with the ImageNet mean and standard deviation.
## Evaluation results
For evaluation results, we refer readers to the [paper](https://arxiv.org/abs/2205.14100). | null |
Salesforce/instructblip-flan-t5-xl |
# InstructBLIP model
InstructBLIP model using [Flan-T5-xl](https://huggingface.co/google/flan-t5-xl) as language model. InstructBLIP was introduced in the paper [InstructBLIP: Towards General-purpose Vision-Language Models with Instruction Tuning](https://arxiv.org/abs/2305.06500) by Dai et al.
Disclaimer: The team releasing InstructBLIP did not write a model card for this model so this model card has been written by the Hugging Face team.
## Model description
InstructBLIP is a visual instruction tuned version of [BLIP-2](https://huggingface.co/docs/transformers/main/model_doc/blip-2). Refer to the paper for details.
## Intended uses & limitations
Usage is as follows:
```
from transformers import InstructBlipProcessor, InstructBlipForConditionalGeneration
import torch
from PIL import Image
import requests
model = InstructBlipForConditionalGeneration.from_pretrained("Salesforce/instructblip-flan-t5-xl")
processor = InstructBlipProcessor.from_pretrained("Salesforce/instructblip-flan-t5-xl")
device = "cuda" if torch.cuda.is_available() else "cpu"
model.to(device)
url = "https://raw.githubusercontent.com/salesforce/LAVIS/main/docs/_static/Confusing-Pictures.jpg"
image = Image.open(requests.get(url, stream=True).raw).convert("RGB")
prompt = "What is unusual about this image?"
inputs = processor(images=image, text=prompt, return_tensors="pt").to(device)
outputs = model.generate(
**inputs,
do_sample=False,
num_beams=5,
max_length=256,
min_length=1,
top_p=0.9,
repetition_penalty=1.5,
length_penalty=1.0,
temperature=1,
)
generated_text = processor.batch_decode(outputs, skip_special_tokens=True)[0].strip()
print(generated_text)
```
### How to use
For code examples, we refer to the [documentation](https://huggingface.co/docs/transformers/main/en/model_doc/instructblip).
## Ethical Considerations
This release is for research purposes only in support of an academic paper. Our models, datasets, and code are not specifically designed or evaluated for all downstream purposes. We strongly recommend users evaluate and address potential concerns related to accuracy, safety, and fairness before deploying this model. We encourage users to consider the common limitations of AI, comply with applicable laws, and leverage best practices when selecting use cases, particularly for high-risk scenarios where errors or misuse could significantly impact people’s lives, rights, or safety. For further guidance on use cases, refer to our AUP and AI AUP.
| null |
noamrot/FuseCap_Image_Captioning | # FuseCap: Leveraging Large Language Models for Enriched Fused Image Captions
A framework designed to generate semantically rich image captions.
## Resources
- 💻 **Project Page**: For more details, visit the official [project page](https://rotsteinnoam.github.io/FuseCap/).
- 📝 **Read the Paper**: You can find the paper [here](https://arxiv.org/abs/2305.17718).
- 🚀 **Demo**: Try out our BLIP-based model [demo](https://huggingface.co/spaces/noamrot/FuseCap) trained using FuseCap.
- 📂 **Code Repository**: The code for FuseCap can be found in the [GitHub repository](https://github.com/RotsteinNoam/FuseCap).
- 🗃️ **Datasets**: The fused captions datasets can be accessed from [here](https://github.com/RotsteinNoam/FuseCap#datasets).
#### Running the model
Our BLIP-based model can be run using the following code,
```python
import requests
from PIL import Image
from transformers import BlipProcessor, BlipForConditionalGeneration
import torch
device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')
processor = BlipProcessor.from_pretrained("noamrot/FuseCap")
model = BlipForConditionalGeneration.from_pretrained("noamrot/FuseCap").to(device)
img_url = 'https://huggingface.co/spaces/noamrot/FuseCap/resolve/main/bike.jpg'
raw_image = Image.open(requests.get(img_url, stream=True).raw).convert('RGB')
text = "a picture of "
inputs = processor(raw_image, text, return_tensors="pt").to(device)
out = model.generate(**inputs, num_beams = 3)
print(processor.decode(out[0], skip_special_tokens=True))
```
## Upcoming Updates
The official codebase, datasets and trained models for this project will be released soon.
## BibTeX
``` Citation
@inproceedings{rotstein2024fusecap,
title={Fusecap: Leveraging large language models for enriched fused image captions},
author={Rotstein, Noam and Bensa{\"\i}d, David and Brody, Shaked and Ganz, Roy and Kimmel, Ron},
booktitle={Proceedings of the IEEE/CVF Winter Conference on Applications of Computer Vision},
pages={5689--5700},
year={2024}
}
``` | null |
Salesforce/instructblip-flan-t5-xxl |
# InstructBLIP model
InstructBLIP model using [Flan-T5-xxl](https://huggingface.co/google/flan-t5-xxl) as language model. InstructBLIP was introduced in the paper [InstructBLIP: Towards General-purpose Vision-Language Models with Instruction Tuning](https://arxiv.org/abs/2305.06500) by Dai et al.
Disclaimer: The team releasing InstructBLIP did not write a model card for this model so this model card has been written by the Hugging Face team.
## Model description
InstructBLIP is a visual instruction tuned version of [BLIP-2](https://huggingface.co/docs/transformers/main/model_doc/blip-2). Refer to the paper for details.
## Intended uses & limitations
Usage is as follows:
```
from transformers import InstructBlipProcessor, InstructBlipForConditionalGeneration
import torch
from PIL import Image
import requests
model = InstructBlipForConditionalGeneration.from_pretrained("Salesforce/instructblip-flan-t5-xxl")
processor = InstructBlipProcessor.from_pretrained("Salesforce/instructblip-flan-t5-xxl")
device = "cuda" if torch.cuda.is_available() else "cpu"
model.to(device)
url = "https://raw.githubusercontent.com/salesforce/LAVIS/main/docs/_static/Confusing-Pictures.jpg"
image = Image.open(requests.get(url, stream=True).raw).convert("RGB")
prompt = "What is unusual about this image?"
inputs = processor(images=image, text=prompt, return_tensors="pt").to(device)
outputs = model.generate(
**inputs,
do_sample=False,
num_beams=5,
max_length=256,
min_length=1,
top_p=0.9,
repetition_penalty=1.5,
length_penalty=1.0,
temperature=1,
)
generated_text = processor.batch_decode(outputs, skip_special_tokens=True)[0].strip()
print(generated_text)
```
## Ethical Considerations
This release is for research purposes only in support of an academic paper. Our models, datasets, and code are not specifically designed or evaluated for all downstream purposes. We strongly recommend users evaluate and address potential concerns related to accuracy, safety, and fairness before deploying this model. We encourage users to consider the common limitations of AI, comply with applicable laws, and leverage best practices when selecting use cases, particularly for high-risk scenarios where errors or misuse could significantly impact people’s lives, rights, or safety. For further guidance on use cases, refer to our AUP and AI AUP.
### How to use
For code examples, we refer to the [documentation](https://huggingface.co/docs/transformers/main/en/model_doc/instructblip). | null |
Salesforce/instructblip-vicuna-13b |
# InstructBLIP model
InstructBLIP model using [Vicuna-13b](https://github.com/lm-sys/FastChat#model-weights) as language model. InstructBLIP was introduced in the paper [InstructBLIP: Towards General-purpose Vision-Language Models with Instruction Tuning](https://arxiv.org/abs/2305.06500) by Dai et al.
Disclaimer: The team releasing InstructBLIP did not write a model card for this model so this model card has been written by the Hugging Face team.
## Model description
InstructBLIP is a visual instruction tuned version of [BLIP-2](https://huggingface.co/docs/transformers/main/model_doc/blip-2). Refer to the paper for details.
## Intended uses & limitations
Usage is as follows:
```
from transformers import InstructBlipProcessor, InstructBlipForConditionalGeneration
import torch
from PIL import Image
import requests
model = InstructBlipForConditionalGeneration.from_pretrained("Salesforce/instructblip-vicuna-13b")
processor = InstructBlipProcessor.from_pretrained("Salesforce/instructblip-vicuna-13b")
device = "cuda" if torch.cuda.is_available() else "cpu"
model.to(device)
url = "https://raw.githubusercontent.com/salesforce/LAVIS/main/docs/_static/Confusing-Pictures.jpg"
image = Image.open(requests.get(url, stream=True).raw).convert("RGB")
prompt = "What is unusual about this image?"
inputs = processor(images=image, text=prompt, return_tensors="pt").to(device)
outputs = model.generate(
**inputs,
do_sample=False,
num_beams=5,
max_length=256,
min_length=1,
top_p=0.9,
repetition_penalty=1.5,
length_penalty=1.0,
temperature=1,
)
generated_text = processor.batch_decode(outputs, skip_special_tokens=True)[0].strip()
print(generated_text)
```
### How to use
For code examples, we refer to the [documentation](https://huggingface.co/docs/transformers/main/en/model_doc/instructblip).
## Ethical Considerations
This release is for research purposes only in support of an academic paper. Our models, datasets, and code are not specifically designed or evaluated for all downstream purposes. We strongly recommend users evaluate and address potential concerns related to accuracy, safety, and fairness before deploying this model. We encourage users to consider the common limitations of AI, comply with applicable laws, and leverage best practices when selecting use cases, particularly for high-risk scenarios where errors or misuse could significantly impact people’s lives, rights, or safety. For further guidance on use cases, refer to our AUP and AI AUP.
| null |
muhualing/vit |
# nlpconnect/vit-gpt2-image-captioning
This is an image captioning model trained by @ydshieh in [flax ](https://github.com/huggingface/transformers/tree/main/examples/flax/image-captioning) this is pytorch version of [this](https://huggingface.co/ydshieh/vit-gpt2-coco-en-ckpts).
# The Illustrated Image Captioning using transformers
* https://ankur3107.github.io/blogs/the-illustrated-image-captioning-using-transformers/
# Sample running code
```python
from transformers import VisionEncoderDecoderModel, ViTImageProcessor, AutoTokenizer
import torch
from PIL import Image
model = VisionEncoderDecoderModel.from_pretrained("nlpconnect/vit-gpt2-image-captioning")
feature_extractor = ViTImageProcessor.from_pretrained("nlpconnect/vit-gpt2-image-captioning")
tokenizer = AutoTokenizer.from_pretrained("nlpconnect/vit-gpt2-image-captioning")
device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
model.to(device)
max_length = 16
num_beams = 4
gen_kwargs = {"max_length": max_length, "num_beams": num_beams}
def predict_step(image_paths):
images = []
for image_path in image_paths:
i_image = Image.open(image_path)
if i_image.mode != "RGB":
i_image = i_image.convert(mode="RGB")
images.append(i_image)
pixel_values = feature_extractor(images=images, return_tensors="pt").pixel_values
pixel_values = pixel_values.to(device)
output_ids = model.generate(pixel_values, **gen_kwargs)
preds = tokenizer.batch_decode(output_ids, skip_special_tokens=True)
preds = [pred.strip() for pred in preds]
return preds
predict_step(['doctor.e16ba4e4.jpg']) # ['a woman in a hospital bed with a woman in a hospital bed']
```
# Sample running code using transformers pipeline
```python
from transformers import pipeline
image_to_text = pipeline("image-to-text", model="nlpconnect/vit-gpt2-image-captioning")
image_to_text("https://ankur3107.github.io/assets/images/image-captioning-example.png")
# [{'generated_text': 'a soccer game with a player jumping to catch the ball '}]
```
# Contact for any help
* https://huggingface.co/ankur310794
* https://twitter.com/ankur310794
* http://github.com/ankur3107
* https://www.linkedin.com/in/ankur310794 | null |
paragon-AI/blip2-image-to-text |
# BLIP-2, OPT-2.7b, pre-trained only
BLIP-2 model, leveraging [OPT-2.7b](https://huggingface.co/facebook/opt-2.7b) (a large language model with 2.7 billion parameters).
It was introduced in the paper [BLIP-2: Bootstrapping Language-Image Pre-training with Frozen Image Encoders and Large Language Models](https://arxiv.org/abs/2301.12597) by Li et al. and first released in [this repository](https://github.com/salesforce/LAVIS/tree/main/projects/blip2).
Disclaimer: The team releasing BLIP-2 did not write a model card for this model so this model card has been written by the Hugging Face team.
## Model description
BLIP-2 consists of 3 models: a CLIP-like image encoder, a Querying Transformer (Q-Former) and a large language model.
The authors initialize the weights of the image encoder and large language model from pre-trained checkpoints and keep them frozen
while training the Querying Transformer, which is a BERT-like Transformer encoder that maps a set of "query tokens" to query embeddings,
which bridge the gap between the embedding space of the image encoder and the large language model.
The goal for the model is simply to predict the next text token, giving the query embeddings and the previous text.
<img src="https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/transformers/model_doc/blip2_architecture.jpg"
alt="drawing" width="600"/>
This allows the model to be used for tasks like:
- image captioning
- visual question answering (VQA)
- chat-like conversations by feeding the image and the previous conversation as prompt to the model
## Direct Use and Downstream Use
You can use the raw model for conditional text generation given an image and optional text. See the [model hub](https://huggingface.co/models?search=Salesforce/blip) to look for
fine-tuned versions on a task that interests you.
## Bias, Risks, Limitations, and Ethical Considerations
BLIP2-OPT uses off-the-shelf OPT as the language model. It inherits the same risks and limitations as mentioned in Meta's model card.
> Like other large language models for which the diversity (or lack thereof) of training
> data induces downstream impact on the quality of our model, OPT-175B has limitations in terms
> of bias and safety. OPT-175B can also have quality issues in terms of generation diversity and
> hallucination. In general, OPT-175B is not immune from the plethora of issues that plague modern
> large language models.
>
BLIP2 is fine-tuned on image-text datasets (e.g. [LAION](https://laion.ai/blog/laion-400-open-dataset/) ) collected from the internet. As a result the model itself is potentially vulnerable to generating equivalently inappropriate content or replicating inherent biases in the underlying data.
BLIP2 has not been tested in real world applications. It should not be directly deployed in any applications. Researchers should first carefully assess the safety and fairness of the model in relation to the specific context they’re being deployed within.
### How to use
For code examples, we refer to the [documentation](https://huggingface.co/docs/transformers/main/en/model_doc/blip-2#transformers.Blip2ForConditionalGeneration.forward.example).
#### Running the model on CPU
<details>
<summary> Click to expand </summary>
```python
import requests
from PIL import Image
from transformers import Blip2Processor, Blip2ForConditionalGeneration
processor = Blip2Processor.from_pretrained("Salesforce/blip2-opt-2.7b")
model = Blip2ForConditionalGeneration.from_pretrained("Salesforce/blip2-opt-2.7b")
img_url = 'https://storage.googleapis.com/sfr-vision-language-research/BLIP/demo.jpg'
raw_image = Image.open(requests.get(img_url, stream=True).raw).convert('RGB')
question = "how many dogs are in the picture?"
inputs = processor(raw_image, question, return_tensors="pt")
out = model.generate(**inputs)
print(processor.decode(out[0], skip_special_tokens=True))
```
</details>
#### Running the model on GPU
##### In full precision
<details>
<summary> Click to expand </summary>
```python
# pip install accelerate
import requests
from PIL import Image
from transformers import Blip2Processor, Blip2ForConditionalGeneration
processor = Blip2Processor.from_pretrained("Salesforce/blip2-opt-2.7b")
model = Blip2ForConditionalGeneration.from_pretrained("Salesforce/blip2-opt-2.7b", device_map="auto")
img_url = 'https://storage.googleapis.com/sfr-vision-language-research/BLIP/demo.jpg'
raw_image = Image.open(requests.get(img_url, stream=True).raw).convert('RGB')
question = "how many dogs are in the picture?"
inputs = processor(raw_image, question, return_tensors="pt").to("cuda")
out = model.generate(**inputs)
print(processor.decode(out[0], skip_special_tokens=True))
```
</details>
##### In half precision (`float16`)
<details>
<summary> Click to expand </summary>
```python
# pip install accelerate
import torch
import requests
from PIL import Image
from transformers import Blip2Processor, Blip2ForConditionalGeneration
processor = Blip2Processor.from_pretrained("Salesforce/blip2-opt-2.7b")
model = Blip2ForConditionalGeneration.from_pretrained("Salesforce/blip2-opt-2.7b", torch_dtype=torch.float16, device_map="auto")
img_url = 'https://storage.googleapis.com/sfr-vision-language-research/BLIP/demo.jpg'
raw_image = Image.open(requests.get(img_url, stream=True).raw).convert('RGB')
question = "how many dogs are in the picture?"
inputs = processor(raw_image, question, return_tensors="pt").to("cuda", torch.float16)
out = model.generate(**inputs)
print(processor.decode(out[0], skip_special_tokens=True))
```
</details>
##### In 8-bit precision (`int8`)
<details>
<summary> Click to expand </summary>
```python
# pip install accelerate bitsandbytes
import torch
import requests
from PIL import Image
from transformers import Blip2Processor, Blip2ForConditionalGeneration
processor = Blip2Processor.from_pretrained("Salesforce/blip2-opt-2.7b")
model = Blip2ForConditionalGeneration.from_pretrained("Salesforce/blip2-opt-2.7b", load_in_8bit=True, device_map="auto")
img_url = 'https://storage.googleapis.com/sfr-vision-language-research/BLIP/demo.jpg'
raw_image = Image.open(requests.get(img_url, stream=True).raw).convert('RGB')
question = "how many dogs are in the picture?"
inputs = processor(raw_image, question, return_tensors="pt").to("cuda", torch.float16)
out = model.generate(**inputs)
print(processor.decode(out[0], skip_special_tokens=True))
```
</details> | null |
captioner/caption-gen | null |
|
movementso/blip-image-captioning-large |
# BLIP: Bootstrapping Language-Image Pre-training for Unified Vision-Language Understanding and Generation
Model card for image captioning pretrained on COCO dataset - base architecture (with ViT large backbone).
|  |
|:--:|
| <b> Pull figure from BLIP official repo | Image source: https://github.com/salesforce/BLIP </b>|
## TL;DR
Authors from the [paper](https://arxiv.org/abs/2201.12086) write in the abstract:
*Vision-Language Pre-training (VLP) has advanced the performance for many vision-language tasks. However, most existing pre-trained models only excel in either understanding-based tasks or generation-based tasks. Furthermore, performance improvement has been largely achieved by scaling up the dataset with noisy image-text pairs collected from the web, which is a suboptimal source of supervision. In this paper, we propose BLIP, a new VLP framework which transfers flexibly to both vision-language understanding and generation tasks. BLIP effectively utilizes the noisy web data by bootstrapping the captions, where a captioner generates synthetic captions and a filter removes the noisy ones. We achieve state-of-the-art results on a wide range of vision-language tasks, such as image-text retrieval (+2.7% in average recall@1), image captioning (+2.8% in CIDEr), and VQA (+1.6% in VQA score). BLIP also demonstrates strong generalization ability when directly transferred to videolanguage tasks in a zero-shot manner. Code, models, and datasets are released.*
## Usage
You can use this model for conditional and un-conditional image captioning
### Using the Pytorch model
#### Running the model on CPU
<details>
<summary> Click to expand </summary>
```python
import requests
from PIL import Image
from transformers import BlipProcessor, BlipForConditionalGeneration
processor = BlipProcessor.from_pretrained("Salesforce/blip-image-captioning-large")
model = BlipForConditionalGeneration.from_pretrained("Salesforce/blip-image-captioning-large")
img_url = 'https://storage.googleapis.com/sfr-vision-language-research/BLIP/demo.jpg'
raw_image = Image.open(requests.get(img_url, stream=True).raw).convert('RGB')
# conditional image captioning
text = "a photography of"
inputs = processor(raw_image, text, return_tensors="pt")
out = model.generate(**inputs)
print(processor.decode(out[0], skip_special_tokens=True))
# unconditional image captioning
inputs = processor(raw_image, return_tensors="pt")
out = model.generate(**inputs)
print(processor.decode(out[0], skip_special_tokens=True))
```
</details>
#### Running the model on GPU
##### In full precision
<details>
<summary> Click to expand </summary>
```python
import requests
from PIL import Image
from transformers import BlipProcessor, BlipForConditionalGeneration
processor = BlipProcessor.from_pretrained("Salesforce/blip-image-captioning-large")
model = BlipForConditionalGeneration.from_pretrained("Salesforce/blip-image-captioning-large").to("cuda")
img_url = 'https://storage.googleapis.com/sfr-vision-language-research/BLIP/demo.jpg'
raw_image = Image.open(requests.get(img_url, stream=True).raw).convert('RGB')
# conditional image captioning
text = "a photography of"
inputs = processor(raw_image, text, return_tensors="pt").to("cuda")
out = model.generate(**inputs)
print(processor.decode(out[0], skip_special_tokens=True))
# unconditional image captioning
inputs = processor(raw_image, return_tensors="pt").to("cuda")
out = model.generate(**inputs)
print(processor.decode(out[0], skip_special_tokens=True))
```
</details>
##### In half precision (`float16`)
<details>
<summary> Click to expand </summary>
```python
import torch
import requests
from PIL import Image
from transformers import BlipProcessor, BlipForConditionalGeneration
processor = BlipProcessor.from_pretrained("Salesforce/blip-image-captioning-large")
model = BlipForConditionalGeneration.from_pretrained("Salesforce/blip-image-captioning-large", torch_dtype=torch.float16).to("cuda")
img_url = 'https://storage.googleapis.com/sfr-vision-language-research/BLIP/demo.jpg'
raw_image = Image.open(requests.get(img_url, stream=True).raw).convert('RGB')
# conditional image captioning
text = "a photography of"
inputs = processor(raw_image, text, return_tensors="pt").to("cuda", torch.float16)
out = model.generate(**inputs)
print(processor.decode(out[0], skip_special_tokens=True))
# >>> a photography of a woman and her dog
# unconditional image captioning
inputs = processor(raw_image, return_tensors="pt").to("cuda", torch.float16)
out = model.generate(**inputs)
print(processor.decode(out[0], skip_special_tokens=True))
>>> a woman sitting on the beach with her dog
```
</details>
## BibTex and citation info
```
@misc{https://doi.org/10.48550/arxiv.2201.12086,
doi = {10.48550/ARXIV.2201.12086},
url = {https://arxiv.org/abs/2201.12086},
author = {Li, Junnan and Li, Dongxu and Xiong, Caiming and Hoi, Steven},
keywords = {Computer Vision and Pattern Recognition (cs.CV), FOS: Computer and information sciences, FOS: Computer and information sciences},
title = {BLIP: Bootstrapping Language-Image Pre-training for Unified Vision-Language Understanding and Generation},
publisher = {arXiv},
year = {2022},
copyright = {Creative Commons Attribution 4.0 International}
}
``` | null |
trojblue/blip2-opt-6.7b-coco-fp16 |
# BLIP-2, OPT-6.7b, Fine-tuned on COCO - Unofficial FP16 Version
This repository contains an unofficial version of the BLIP-2 model, leveraging [OPT-6.7b](https://huggingface.co/facebook/opt-6.7b), which has been fine-tuned on COCO and converted to FP16 for reduced model size and memory footprint.
The original model, BLIP-2, was introduced in the paper [BLIP-2: Bootstrapping Language-Image Pre-training with Frozen Image Encoders and Large Language Models](https://arxiv.org/abs/2301.12597) by Li et al. and first released in [this repository](https://github.com/salesforce/LAVIS/tree/main/projects/blip2).
For a comprehensive understanding of the model, its description, intended uses, limitations, and instructions on usage with different hardware and precision settings, please refer to the [official model card](https://huggingface.co/Salesforce/blip2-opt-6.7b-coco).
## Unofficial FP16 Version
This version of the BLIP-2 model has been converted to use FP16 precision, which effectively reduces the model size and memory requirements. The conversion to FP16 can potentially accelerate the model's computation time on hardware with FP16 support, although it might slightly affect the model's performance due to reduced numerical precision.
This unofficial FP16 version is ideal for situations where storage, memory, or computational resources are limited.
Please note, this is an **unofficial** repository and not maintained or endorsed by the original authors of the model. The FP16 conversion was conducted independently and any potential issues, limitations or discrepancies with the original model are not the responsibility of the original authors.
### How to use
The usage of this FP16 version of the model is similar to the original model. For specific code examples, we refer to the [documentation](https://huggingface.co/docs/transformers/main/en/model_doc/blip-2#transformers.Blip2ForConditionalGeneration.forward.example).
Please ensure to test the performance and accuracy of this FP16 model thoroughly in your specific use-case to confirm it meets your needs.
This version can be used for tasks like:
- image captioning
- visual question answering (VQA)
- chat-like conversations by feeding the image and the previous conversation as a prompt to the model
*Disclaimer: This is an unofficial version of the model and any potential issues or discrepancies from the official model are not the responsibility of the original authors.* | null |
LanguageMachines/blip2-flan-t5-xxl |
# BLIP-2, Flan T5-xxl, pre-trained only
BLIP-2 model, leveraging [Flan T5-xxl](https://huggingface.co/google/flan-t5-xxl) (a large language model).
It was introduced in the paper [BLIP-2: Bootstrapping Language-Image Pre-training with Frozen Image Encoders and Large Language Models](https://arxiv.org/abs/2301.12597) by Li et al. and first released in [this repository](https://github.com/salesforce/LAVIS/tree/main/projects/blip2).
Disclaimer: The team releasing BLIP-2 did not write a model card for this model so this model card has been written by the Hugging Face team.
## Model description
BLIP-2 consists of 3 models: a CLIP-like image encoder, a Querying Transformer (Q-Former) and a large language model.
The authors initialize the weights of the image encoder and large language model from pre-trained checkpoints and keep them frozen
while training the Querying Transformer, which is a BERT-like Transformer encoder that maps a set of "query tokens" to query embeddings,
which bridge the gap between the embedding space of the image encoder and the large language model.
The goal for the model is simply to predict the next text token, giving the query embeddings and the previous text.
<img src="https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/transformers/model_doc/blip2_architecture.jpg"
alt="drawing" width="600"/>
This allows the model to be used for tasks like:
- image captioning
- visual question answering (VQA)
- chat-like conversations by feeding the image and the previous conversation as prompt to the model
## Direct Use and Downstream Use
You can use the raw model for conditional text generation given an image and optional text. See the [model hub](https://huggingface.co/models?search=Salesforce/blip) to look for
fine-tuned versions on a task that interests you.
## Bias, Risks, Limitations, and Ethical Considerations
BLIP2-FlanT5 uses off-the-shelf Flan-T5 as the language model. It inherits the same risks and limitations from [Flan-T5](https://arxiv.org/pdf/2210.11416.pdf):
> Language models, including Flan-T5, can potentially be used for language generation in a harmful way, according to Rae et al. (2021). Flan-T5 should not be used directly in any application, without a prior assessment of safety and fairness concerns specific to the application.
BLIP2 is fine-tuned on image-text datasets (e.g. [LAION](https://laion.ai/blog/laion-400-open-dataset/) ) collected from the internet. As a result the model itself is potentially vulnerable to generating equivalently inappropriate content or replicating inherent biases in the underlying data.
BLIP2 has not been tested in real world applications. It should not be directly deployed in any applications. Researchers should first carefully assess the safety and fairness of the model in relation to the specific context they’re being deployed within.
### How to use
For code examples, we refer to the [documentation](https://huggingface.co/docs/transformers/main/en/model_doc/blip-2#transformers.Blip2ForConditionalGeneration.forward.example), or refer to the snippets below depending on your usecase:
#### Running the model on CPU
<details>
<summary> Click to expand </summary>
```python
import requests
from PIL import Image
from transformers import BlipProcessor, Blip2ForConditionalGeneration
processor = BlipProcessor.from_pretrained("Salesforce/blip2-flan-t5-xxl")
model = Blip2ForConditionalGeneration.from_pretrained("Salesforce/blip2-flan-t5-xxl")
img_url = 'https://storage.googleapis.com/sfr-vision-language-research/BLIP/demo.jpg'
raw_image = Image.open(requests.get(img_url, stream=True).raw).convert('RGB')
question = "how many dogs are in the picture?"
inputs = processor(raw_image, question, return_tensors="pt")
out = model.generate(**inputs)
print(processor.decode(out[0], skip_special_tokens=True))
```
</details>
#### Running the model on GPU
##### In full precision
<details>
<summary> Click to expand </summary>
```python
# pip install accelerate
import requests
from PIL import Image
from transformers import Blip2Processor, Blip2ForConditionalGeneration
processor = Blip2Processor.from_pretrained("Salesforce/blip2-flan-t5-xxl")
model = Blip2ForConditionalGeneration.from_pretrained("Salesforce/blip2-flan-t5-xxl", device_map="auto")
img_url = 'https://storage.googleapis.com/sfr-vision-language-research/BLIP/demo.jpg'
raw_image = Image.open(requests.get(img_url, stream=True).raw).convert('RGB')
question = "how many dogs are in the picture?"
inputs = processor(raw_image, question, return_tensors="pt").to("cuda")
out = model.generate(**inputs)
print(processor.decode(out[0], skip_special_tokens=True))
```
</details>
##### In half precision (`float16`)
<details>
<summary> Click to expand </summary>
```python
# pip install accelerate
import torch
import requests
from PIL import Image
from transformers import Blip2Processor, Blip2ForConditionalGeneration
processor = Blip2Processor.from_pretrained("Salesforce/blip2-flan-t5-xxl")
model = Blip2ForConditionalGeneration.from_pretrained("Salesforce/blip2-flan-t5-xxl", torch_dtype=torch.float16, device_map="auto")
img_url = 'https://storage.googleapis.com/sfr-vision-language-research/BLIP/demo.jpg'
raw_image = Image.open(requests.get(img_url, stream=True).raw).convert('RGB')
question = "how many dogs are in the picture?"
inputs = processor(raw_image, question, return_tensors="pt").to("cuda", torch.float16)
out = model.generate(**inputs)
print(processor.decode(out[0], skip_special_tokens=True))
```
</details>
##### In 8-bit precision (`int8`)
<details>
<summary> Click to expand </summary>
```python
# pip install accelerate bitsandbytes
import torch
import requests
from PIL import Image
from transformers import Blip2Processor, Blip2ForConditionalGeneration
processor = Blip2Processor.from_pretrained("Salesforce/blip2-flan-t5-xxl")
model = Blip2ForConditionalGeneration.from_pretrained("Salesforce/blip2-flan-t5-xxl", load_in_8bit=True, device_map="auto")
img_url = 'https://storage.googleapis.com/sfr-vision-language-research/BLIP/demo.jpg'
raw_image = Image.open(requests.get(img_url, stream=True).raw).convert('RGB')
question = "how many dogs are in the picture?"
inputs = processor(raw_image, question, return_tensors="pt").to("cuda", torch.float16)
out = model.generate(**inputs)
print(processor.decode(out[0], skip_special_tokens=True))
```
</details> | null |
End of preview. Expand
in Data Studio
README.md exists but content is empty.
- Downloads last month
- 34
Size of downloaded dataset files:
150 kB
Size of the auto-converted Parquet files:
150 kB
Number of rows:
179