ctmedtech/DDR-dataset

DDR - Diabetic Retinopathy Detection Dataset

Image: Dataset Samples.

---

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
• Source: Kaggle - DDR Dataset
• 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

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.