Shivakarthik12/melanoma-classifier

Melanoma classifier

This Space packages the existing notebook inference path. It uses the checkpoint metadata for the Swin V2 model configuration, class names, preprocessing, and threshold.

The notebook's saved single-image inference function is deterministic Resize -> Normalize -> ToTensorV2, followed by one model forward pass and thresholding of the melanoma probability. Its evaluation-only TTA/calibration experiments are not applied here because they are not part of that current single-image inference function.

Upload best_swin_checkpoint.pth to this repository. The checkpoint is intentionally not converted or replaced.