Diabetic Retinopathy Detection System
A full-stack AI-powered web application for automated Diabetic Retinopathy grading from retinal fundus photographs.
Model
Architecture: EfficientDRNet — Custom CNN with CBAM (Channel + Spatial Attention) Dataset: APTOS 2019 Blindness Detection Classes: 0 — No DR, 1 — Mild, 2 — Moderate, 3 — Severe Metrics: Accuracy 83.3% | QWK 0.9184 | AUC 0.9546 Explainability: Grad-CAM heatmap visualization