eel4759_classification/utils.py, branch master Comparison of different ImageNet-based CNN models for classifying diabetic retinopathy images improve training for class-imbalanced DR classification 2026-04-20T14:03:17+00:00 Vineet Kumar git@vineetk.net 2026-04-20T14:03:17+00:00 9dd8b3007e79f0634f868d405f5e7c48917305fd - WeightedRandomSampler for balanced class sampling during training - FocalLoss (gamma=2) for hard-example mining, no alpha (sampler handles balance) - Stronger augmentation: RandomResizedCrop, ColorJitter, GaussianBlur, 30deg rotation - Discriminative LR: backbone at 0.1x, head at 1x - 3-epoch linear warmup + cosine decay scheduler - Early stopping on macro F1 instead of weighted F1 - Updated defaults: batch_size=64, epochs=50, patience=10 - Comparison chart now shows both macro and weighted F1 - DataParallel checkpoint handling (strip module. prefix) - Enable cudnn.benchmark for training speed 
- WeightedRandomSampler for balanced class sampling during training
- FocalLoss (gamma=2) for hard-example mining, no alpha (sampler handles balance)
- Stronger augmentation: RandomResizedCrop, ColorJitter, GaussianBlur, 30deg rotation
- Discriminative LR: backbone at 0.1x, head at 1x
- 3-epoch linear warmup + cosine decay scheduler
- Early stopping on macro F1 instead of weighted F1
- Updated defaults: batch_size=64, epochs=50, patience=10
- Comparison chart now shows both macro and weighted F1
- DataParallel checkpoint handling (strip module. prefix)
- Enable cudnn.benchmark for training speed
initial commit 2026-04-19T01:34:11+00:00 Vineet Kumar git@vineetk.net 2026-04-19T01:34:11+00:00 8f571ee51854540d186aafc01d25908bba5ea09e