---
license: apache-2.0
task_categories:
- object-detection
tags:
- roboflow
- signature
pretty_name: Handwritten Signature
size_categories:
- 1K<n<10K
configs:
- config_name: full
  data_files:
  - split: train
    path: full/train-*
  - split: validation
    path: full/validation-*
  - split: test
    path: full/test-*
  default: true
dataset_info:
  config_name: full
  features:
  - name: image_id
    dtype: int64
  - name: image
    dtype: image
  - name: width
    dtype: int32
  - name: height
    dtype: int32
  - name: objects
    sequence:
    - name: id
      dtype: int64
    - name: area
      dtype: int64
    - name: bbox
      sequence: float32
      length: 4
    - name: category
      dtype:
        class_label:
          names:
            '0': signature
  splits:
  - name: train
    num_bytes: 114346924.72
    num_examples: 1980
  - name: validation
    num_bytes: 18085018
    num_examples: 420
  - name: test
    num_bytes: 18307713
    num_examples: 419
  download_size: 146763157
  dataset_size: 150739655.72
---

# **Dataset: Signature Detection**

This dataset was developed to train models for handwritten signature detection in various types of documents. It combines data from two public datasets ([Tobacco800](https://paperswithcode.com/dataset/tobacco-800) and [signatures-xc8up](https://universe.roboflow.com/roboflow-100/signatures-xc8up)) with processing and unification performed in [Roboflow](https://roboflow.com/).

![](./assets/roboflow_ds.png)

## **Project Resources Overview**

| Resource                        | Links / Badges                                                                                                                                                                                                                                                                                                                   | Details                                                                                                                                                                 |
|---------------------------------|----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------|--------------------------------------------------------------------------------------------------------------------------------------------------------------------------|
| **Article** | [![Paper page](https://huggingface.co/datasets/huggingface/badges/resolve/main/paper-page-md.svg)](https://huggingface.co/blog/samuellimabraz/signature-detection-model) | A detailed community article covering the full development process of the project |
| **Model Files**                 | [![HF Model](https://huggingface.co/datasets/huggingface/badges/resolve/main/model-on-hf-md.svg)](https://huggingface.co/tech4humans/yolov8s-signature-detector)                                                                                                                                                             | **Available formats:** [![PyTorch](https://img.shields.io/badge/PyTorch-%23EE4C2C.svg?style=flat&logo=PyTorch&logoColor=white)](https://pytorch.org/) [![ONNX](https://img.shields.io/badge/ONNX-005CED.svg?style=flat&logo=ONNX&logoColor=white)](https://onnx.ai/) [![TensorRT](https://img.shields.io/badge/TensorRT-76B900.svg?style=flat&logo=NVIDIA&logoColor=white)](https://developer.nvidia.com/tensorrt) |
| **Dataset – Original**          | [![Roboflow](https://app.roboflow.com/images/download-dataset-badge.svg)](https://universe.roboflow.com/tech-ysdkk/signature-detection-hlx8j)                                                                                                                                                                          | 2,819 document images annotated with signature coordinates                                                                                                           |
| **Dataset – Processed**         | [![HF Dataset](https://huggingface.co/datasets/huggingface/badges/resolve/main/dataset-on-hf-md.svg)](https://huggingface.co/datasets/tech4humans/signature-detection)                                                                                                                                                  | Augmented and pre-processed version (640px) for model training                                                                                                          |
| **Notebooks – Model Experiments** | [![Colab](https://colab.research.google.com/assets/colab-badge.svg)](https://colab.research.google.com/drive/1wSySw_zwyuv6XSaGmkngI4dwbj-hR4ix) [![W&B Training](https://img.shields.io/badge/W%26B_Training-FFBE00?style=flat&logo=WeightsAndBiases&logoColor=white)](https://api.wandb.ai/links/samuel-lima-tech4humans/30cmrkp8) | Complete training and evaluation pipeline with selection among different architectures (yolo, detr, rt-detr, conditional-detr, yolos)                                        |
| **Notebooks – HP Tuning**       | [![Colab](https://colab.research.google.com/assets/colab-badge.svg)](https://colab.research.google.com/drive/1wSySw_zwyuv6XSaGmkngI4dwbj-hR4ix) [![W&B HP Tuning](https://img.shields.io/badge/W%26B_HP_Tuning-FFBE00?style=flat&logo=WeightsAndBiases&logoColor=white)](https://api.wandb.ai/links/samuel-lima-tech4humans/31a6zhb1) | Optuna trials for optimizing the precision/recall balance                                                                                                               |
| **Inference Server**            | [![GitHub](https://img.shields.io/badge/Deploy-ffffff?style=for-the-badge&logo=github&logoColor=black)](https://github.com/tech4ai/t4ai-signature-detect-server)                                                                                                                                         | Complete deployment and inference pipeline with Triton Inference Server<br> [![OpenVINO](https://img.shields.io/badge/OpenVINO-00c7fd?style=flat&logo=intel&logoColor=white)](https://docs.openvino.ai/2025/index.html) [![Docker](https://img.shields.io/badge/Docker-2496ED?logo=docker&logoColor=fff)](https://www.docker.com/) [![Triton](https://img.shields.io/badge/Triton-Inference%20Server-76B900?labelColor=black&logo=nvidia)](https://developer.nvidia.com/triton-inference-server) |
| **Live Demo**                   | [![HF Space](https://huggingface.co/datasets/huggingface/badges/resolve/main/open-in-hf-spaces-md.svg)](https://huggingface.co/spaces/tech4humans/signature-detection)                                                                                                                                             | Graphical interface with real-time inference<br> [![Gradio](https://img.shields.io/badge/Gradio-FF5722?style=flat&logo=Gradio&logoColor=white)](https://www.gradio.app/) [![Plotly](https://img.shields.io/badge/PLotly-000000?style=flat&logo=plotly&logoColor=white)](https://plotly.com/python/) |

## Dataset Components

1. **[Tobacco800](https://paperswithcode.com/dataset/tobacco-800):**
   - Subset of the Complex Document Image Processing (CDIP) Test Collection.
   - Contains scanned images of documents related to the tobacco industry, created by the Illinois Institute of Technology.

2. **[signatures-xc8up](https://universe.roboflow.com/roboflow-100/signatures-xc8up):**
   - Part of [Roboflow 100](https://rf100.org/), an Intel initiative.
   - Includes 368 annotated images for handwritten signature detection.

Both were unified to provide a robust and diverse foundation for object detection tasks.

### **Dataset Details**

- **Dataset Split:**
  - Training: 1,980 images (70%)
  - Validation: 420 images (15%)
  - Testing: 419 images (15%)

- **Format:** COCO JSON
- **License:** Apache 2.0

### **Preprocessing and Augmentations**

- **Preprocessing:**
  - Auto-Orientation: Applied
  - Resizing: 640x640 pixels

- **Applied Augmentations:**
  - 90° Rotation: Clockwise, counterclockwise, and upside down
  - Rotation: Between -10° and +10°
  - Shearing: ±4° Horizontal, ±3° Vertical
  - Brightness: Between -8% and +8%
  - Exposure: Between -13% and +13%
  - Blur: Up to 1.1 pixels
  - Noise: Up to 0.97% of pixels

These steps were implemented to enhance the model's robustness and generalization ability.

---

## **Model**

This dataset was used to train the [yolov8s-signature-detector](https://huggingface.co/tech4humans/yolov8s-signature-detector) model for handwritten signature detection. For full technical details including performance metrics and architecture specifications, see the [Model Card](https://huggingface.co/tech4humans/yolov8s-signature-detector).

---

## **How to Use with the Datasets Library**

This dataset is available on the Hugging Face Hub and can be loaded directly using the `datasets` library.

### **Installing the Library**

```bash
pip install datasets
```

### **Loading the Dataset** 

```python
from datasets import load_dataset

dataset = load_dataset("samuellimabraz/signature-detection")

# Visualyze the first sample
print(dataset["train"][0])
```

### **Use Case Example**

```python
import matplotlib.pyplot as plt
import matplotlib.patches as patches
import random
from datasets import load_dataset

dataset = load_dataset("samuellimabraz/signature-detection")

# Randomly select a sample from the test set
sample = dataset["test"][random.randint(0, len(dataset["test"]))]

image = sample["image"]
bboxes = sample["objects"]["bbox"]

fig, ax = plt.subplots(1, figsize=(8, 8))
ax.imshow(image)

for bbox in bboxes:
    x, y, width, height = bbox
    rect = patches.Rectangle(
        (x, y), width, height, linewidth=2, edgecolor="red", facecolor="none"
    )
    ax.add_patch(rect)

plt.axis("off")
plt.show()
```

---

## **License**

The dataset is distributed under the [Apache 2.0](https://www.apache.org/licenses/LICENSE-2.0) license. You are free to use, modify, and distribute the dataset as long as you comply with the license terms.

---

## **Contact and Information**

For more information, questions, and contributions, please contact iag@tech4h.com.br.

<div align="center">
  <p>
    📧 <b>Email:</b> <a href="mailto:iag@tech4h.com.br">iag@tech4h.com.br</a><br>
    🌐 <b>Website:</b> <a href="https://www.tech4.ai/">www.tech4.ai</a><br>
    💼 <b>LinkedIn:</b> <a href="https://www.linkedin.com/company/tech4humans-hyperautomation/">Tech4Humans</a>
  </p>
</div>

## **Author**

<div align="center">
  <table>
    <tr>
      <td align="center" width="140">
        <a href="https://huggingface.co/samuellimabraz">
          <img src="https://avatars.githubusercontent.com/u/115582014?s=400&u=c149baf46c51fdee45ad5344cf1b360236d90d09&v=4" width="120" alt="Samuel Lima"/>
          <h3>Samuel Lima</h3>
        </a>
        <p><i>AI Research Engineer</i></p>
        <p>
          <a href="https://huggingface.co/samuellimabraz">
            <img src="https://img.shields.io/badge/🤗_HuggingFace-samuellimabraz-orange" alt="HuggingFace"/>
          </a>
        </p>
      </td>
      <td width="500">
        <h4>Responsibilities in this Project</h4>
        <ul>
          <li>🔬 Model development and training</li>
          <li>📊 Dataset analysis and processing</li>
          <li>⚙️ Hyperparameter optimization and performance evaluation</li>
          <li>📝 Technical documentation and model card</li>
        </ul>
      </td>
    </tr>
  </table>
</div>

---

<div align="center">
  <p>Developed with 💜 by <a href="https://www.tech4.ai/">Tech4Humans</a></p>
</div>