DDR: Diabetic Retinopathy Detection with Image-Level and Pixel-Level Annotations

DDR is a large-scale collection of retinal fundus images for diabetic retinopathy analysis. The dataset provides both image-level DR grades and pixel-level annotations of pathological lesions, supporting multiple computer vision tasks. It was uploaded to Hugging Face by author 'chanchahh' and was last updated on January 4, 2026.

Use Cases

• Train diabetic retinopathy grading models based on the provided image-level DR labels.
• Develop lesion segmentation algorithms based on the pixel-level annotations of pathological features.
• Evaluate multi-task learning approaches using the combined classification and segmentation annotations.

Strengths

• Provides both image-level diagnostic labels and pixel-level segmentation masks, enabling multi-task learning.
• Described as a 'large-scale collection' of retinal fundus images, suggesting substantial size for model training.

Limitations

• Specific dataset size, row count, and file formats are unknown, limiting suitability assessment.
• Column-level documentation and sample data are unavailable, requiring manual inspection after download.

Provenance

Source

huggingface

Freshness

Last updated 2026-01-04 01:30:09; freshness should be verified.