Update README.md
Browse filesadd description for custom-caption part
README.md
CHANGED
|
@@ -109,10 +109,15 @@ sentences = [
|
|
| 109 |
"Create the original colors of this image"
|
| 110 |
]
|
| 111 |
```
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 112 |
|
| 113 |
### Loading the Dataset
|
| 114 |
|
| 115 |
-
You can load this dataset using the Hugging Face `datasets` library:
|
| 116 |
```python
|
| 117 |
from datasets import load_dataset
|
| 118 |
```
|
|
@@ -133,6 +138,14 @@ train_dataset = load_dataset("nickpai/coco2017-colorization", split="train", rev
|
|
| 133 |
val_dataset = load_dataset("nickpai/coco2017-colorization", split="validation", revision="caption-free")
|
| 134 |
```
|
| 135 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 136 |
## Filtering Criteria
|
| 137 |
|
| 138 |
### 1. Grayscale Images
|
|
@@ -147,4 +160,5 @@ val_dataset = load_dataset("nickpai/coco2017-colorization", split="validation",
|
|
| 147 |
- Images with low color variance, determined by a specified threshold, are removed.
|
| 148 |
- Low color variance can indicate poor image quality or uniform color distribution.
|
| 149 |
|
| 150 |
-
For more details about the filtering criteria,
|
|
|
|
|
|
| 109 |
"Create the original colors of this image"
|
| 110 |
]
|
| 111 |
```
|
| 112 |
+
- **custom-caption:** Provides prompts generated by
|
| 113 |
+
[CLIP Interrogator](https://github.com/pharmapsychotic/clip-interrogator/tree/main) with `'ViT-H-14/laion2b_s32b_b79k'` model.
|
| 114 |
+
Then filter with `'csv_filter.py'` to remove unlikely words, such as black and white, monochrome, grainy, desaturated, etc.
|
| 115 |
+
For more details about the prompts filtering criteria,
|
| 116 |
+
refer to the [Dataset-for-Image-Colorization](https://github.com/nick8592/Dataset-for-Image-Colorization.git) repository.
|
| 117 |
|
| 118 |
### Loading the Dataset
|
| 119 |
|
| 120 |
+
You can load this dataset using the Hugging Face `'datasets'` library:
|
| 121 |
```python
|
| 122 |
from datasets import load_dataset
|
| 123 |
```
|
|
|
|
| 138 |
val_dataset = load_dataset("nickpai/coco2017-colorization", split="validation", revision="caption-free")
|
| 139 |
```
|
| 140 |
|
| 141 |
+
#### Custom-Caption Branch
|
| 142 |
+
```python
|
| 143 |
+
# Load the train split of the colorization dataset from the custom-caption branch
|
| 144 |
+
train_dataset = load_dataset("nickpai/coco2017-colorization", split="train", revision="custom-caption")
|
| 145 |
+
# Load the validation split of the colorization dataset from the custom-caption branch
|
| 146 |
+
val_dataset = load_dataset("nickpai/coco2017-colorization", split="validation", revision="custom-caption")
|
| 147 |
+
```
|
| 148 |
+
|
| 149 |
## Filtering Criteria
|
| 150 |
|
| 151 |
### 1. Grayscale Images
|
|
|
|
| 160 |
- Images with low color variance, determined by a specified threshold, are removed.
|
| 161 |
- Low color variance can indicate poor image quality or uniform color distribution.
|
| 162 |
|
| 163 |
+
For more details about the image filtering criteria,
|
| 164 |
+
refer to the [Dataset-for-Image-Colorization](https://github.com/nick8592/Dataset-for-Image-Colorization.git) repository.
|