Surface defect segmentation (DL GenAI T3 2026)
Predicts masks for 4 defect types on 128 x 800 grayscale images.
Three models are saved here: model1_tiny_unet (from scratch), model2_unet_resnet34 and model3_fpn_convnext_small (both pretrained encoders). Each model has a .pth state dict and a config json with the tuned threshold and min area per class.
Best model: model3_fpn_convnext_small
Validation results (score = 0.7 * dice on defects + 0.3 * empty accuracy):
model val_score val_f1 val_acc val_dice val_empty_acc
model3_fpn_convnext_small 0.758879 0.779887 0.930515 0.682151 0.937913 model2_unet_resnet34 0.704880 0.725242 0.900380 0.617530 0.908696 model1_tiny_unet 0.606153 0.565949 0.804125 0.520019 0.807130