Pytorch-UNet icon indicating copy to clipboard operation
Pytorch-UNet copied to clipboard

Test and predicted mask always returns zero

Open ahmetfurkaann opened this issue 1 year ago • 1 comments

I'm doing multiclass segmentation and it returns me the mask with a black value of 0, both in the validation part and in the prediction part.

ahmetfurkaann avatar Aug 21 '23 06:08 ahmetfurkaann

Hello! I have the same issue. Did you manage to solve it @ahmetfurkaann ? How are your input masks formatted? Do the files have a channel for each class, each with 1's or 255's? Or just two channels and a class index for each pixel?

naguileraleal avatar Sep 24 '23 19:09 naguileraleal