aptos-2019

PyTorch/FastAPI diabetic retinopathy screening API that classifies retinal fundus images into five severity grades using an EfficientNet-B0 ensemble trained on APTOS 2019. Includes authenticated inference, scan history, and cross-validation training utilities.

RetinaGuard is a robust medical imaging system designed to diagnose Diabetic Retinopathy using a Dual-ResNet Ensemble (ResNet101 & ResNet152) and Ben Graham's preprocessing method. It optimizes for the Quadratic Weighted Kappa metric to handle variable image quality in rural healthcare settings.

Deep learning-based classification of Diabetic Retinopathy using the APTOS 2019 dataset and a ResNet-152 architecture to detect stages from 'No DR' to 'Proliferative DR'.