forcesh / Konica-Minolta-Pathological-Image-Segmentation-Challenge

Cancer region segmentation for digital pathological images

Konica-Minolta-Pathological-Image-Segmentation-Challenge

Cancer region segmentation for digital pathological images based on U-Net model.

Digital pathological image

Cancer regions(white)

Results: 826,805.58 score on private lb(16 place), 815,042.40 on public lb(9 place)

Datasets:

1. https://drive.google.com/open?id=0B1upW33U2iNWZDdLM3BmQWItYnc
2. https://drive.google.com/open?id=0B1upW33U2iNWZEZRZC1TSVhzRVE

Weights: https://drive.google.com/open?id=0B1upW33U2iNWTXVBOHE3MUx0TmM