amanchandra/cataract-resnet18-classifier
0
Cataract Detection ResNet18 Model
Trained on ~1170 clinical images (binary: cataract vs non-cataract).
Architecture: ResNet18 Image Size: 224x224 Train Accuracy: 100% Test Accuracy: 97.16% Use: Baseline model for severity grading project
Future Steps
- Grad-CAM to interpret focus region (pupil/lens)
- Region-based unsupervised feature extraction
- Clustering for cataract severity grading