Upload final_model_combined.pth
#1
by JAH17 - opened
Combining segmentation model (UNet++ EfficientNet-B4, KFold+TTA) with classification model (MaxViT-L) into a single end-to-end inference model with dual output for skin lesion analysis.
Input : normalized image tensor [B, 3, H, W]
Output : classification logits [B, 7] + segmentation mask [B, 1, H, W]