Built an end-to-end deep learning pipeline using ResNet-50 to classify retinal images into five stages of Diabetic Retinopathy. Applied transfer learning, image preprocessing, and AUC-based evaluation on the APTOS 2019 Kaggle dataset, achieving a 94% validation AUC—offering real-world potential in clinical diagnosis automation.
computer-vision deep-learning tensorflow keras medical-imaging image-classification transfer-learning data-augmentation cnn-model clinical-decision-support multi-class-classification optical-imaging diabetic-retinopathy resnet50 early-detection medical-diagnosis healthcare-ai retinal-image-analysis aptos2019 auc-evaluation
-
Updated
Mar 12, 2026 - Python