jerrodz/DDR-dataset

---
language:
- en
license:
- cc-by-4.0
tags:
- medical
- ophthalmology
- fundus-image
- image-segmentation
- image-classification
- diabetic-retinopathy
- lesion-segmentation
task_categories:
- image-segmentation
- image-classification
- object-detection
task_ids:
- multi-class-image-classification
- semantic-segmentation
pretty_name: DDR (Diabetic Retinopathy Detection) Dataset
size_categories:
- 10K<n<100K
annotations_creators:
- expert-generated
source_datasets:
- original
paperswithcode_id: ddr
source_data_urls:
- https://www.sciencedirect.com/science/article/abs/pii/S0020025519305377
- https://www.kaggle.com/datasets/mariaherrerot/ddrdataset
---

# DDR - Diabetic Retinopathy Detection Dataset

<table align="center">
<tr>
<td width="100%" align="center">
<img src="rm_images/Merged_Fundus_Images_with_Captions.jpg" alt="Merged Dataset Samples" style="max-width: 100%; height: auto;">
<br>
<p><strong>Image:</strong> Dataset Samples.</p>
</td>
</tr>
</table>

---

The **DDR (Diabetic Retinopathy Detection)** dataset is a large-scale collection of retinal fundus images designed for training and evaluating algorithms in **diabetic retinopathy (DR) grading** and **lesion-level segmentation**. It provides both image-level DR labels and pixel-level annotations of pathological features, making it suitable for classification and segmentation tasks.

---

## Dataset Overview

- **Full Name:** Diabetic Retinopathy Detection and Segmentation Dataset (DDR)
- **Authors:** Yuhao Zhang, Mingxia Liu, Qianni Zhang, et al.
- **Associated Paper:**
*Diabetic Retinopathy Lesion Segmentation Method Based on Multi-Scale Attention and Lesion Perception*
Published in *Information Sciences*, Volume 501, 2019.
[ScienceDirect Link](https://www.sciencedirect.com/science/article/abs/pii/S0020025519305377)
- **Source:** [Kaggle - DDR Dataset](https://www.kaggle.com/datasets/mariaherrerot/ddrdataset)
- **Institution:** Chinese Academy of Sciences, Beijing, China
- **License:** CC BY 4.0

---

## Dataset Structure

### 🧩 Categories
The dataset includes five DR severity levels, labeled according to the International Clinical Diabetic Retinopathy (ICDR) scale:

| Label | Description |
|:------|:--------------------------|
| 0 | No Diabetic Retinopathy |
| 1 | Mild Nonproliferative DR |
| 2 | Moderate Nonproliferative DR |
| 3 | Severe Nonproliferative DR |
| 4 | Proliferative DR |

Additionally, lesion masks are provided for:
- Microaneurysms
- Hemorrhages
- Hard exudates
- Soft exudates

---

## Data Summary

| Split | # Images | Annotation Type | Image Resolution |
|:------|:----------|:----------------|:-----------------|
| Train | ~9,000 | Image-level + lesion masks | 3216×2136 px (avg.) |
| Test | ~1,000 | Image-level + lesion masks | 3216×2136 px (avg.) |

Total: ~10,000 color fundus images collected from multiple clinical sites in China.

---

## Applications

- **Diabetic Retinopathy Classification**
- **Lesion Segmentation and Detection**
- **Multi-scale Attention and Lesion-Aware Learning**
- **Retinal Disease Screening Benchmarking**

---

## Example Usage

```python
from datasets import load_dataset

dataset = load_dataset("your-username/ddr-dataset")
example = dataset["train"][0]
image = example["image"]
mask = example["segmentation_mask"]
```

Citation

If you use this dataset, please cite:

Zhang Y, Liu M, Zhang Q, et al.
Diabetic Retinopathy Lesion Segmentation Method Based on Multi-Scale Attention and Lesion Perception.
Information Sciences, 2019; 501: 511–522.
DOI: 10.1016/j.ins.2019.06.016

Acknowledgements

This dataset was originally collected and published by the Chinese Academy of Sciences and released for research use under a CC BY 4.0 license.
Kaggle rehosting by Mariah Herrero.