JS=1,DS>1and loss>1000 - Githubissues

LeeJunHyun / Image_Segmentation

Pytorch implementation of U-Net, R2U-Net, Attention U-Net, and Attention R2U-Net.

2.66k stars 594 forks source link

JS=1,DS>1and loss>1000 #79

Closed kimjisoo12 closed 1 year ago

kimjisoo12 commented 2 years ago

Hello, SORRY to bother you, when I refocused m attention Unet on my custom dataset, I also encountered JS which was always 1 and DICE >1, I saw previous comments, I changed Torch. Max to Max without any improvement, and my loss was always over 1000, what is the reason?

223

Lloyd-Pottiger commented 2 years ago

update the evaluation.py into the following version: https://github.com/LeeJunHyun/Image_Segmentation/blob/35b98018bf7efcb733ec214200c0ebbdec797f28/evaluation.py