search

JunMa11 / SegWithDistMap

How Distance Transform Maps Boost Segmentation CNNs: An Empirical Study
https://openreview.net/forum?id=hM4pNbXWst
Apache License 2.0
375 stars 67 forks source link

About Loss Divergent #19

Closed YangHan-Morningstar closed 2 years ago

YangHan-Morningstar commented 2 years ago

Hi, thanks for your great work! When I run train_LA_AAAISDF.py, I use tensor board to see the change of total loss function and found it became divergent after 1.4k. So I use the model in 1k to test and get an unexpected result, Could tell me how to fix it? Thanks a lot!

image image
JunMa11 commented 2 years ago

Hi @YangHan-Morningstar ,

Thanks for your interest very much.

However, it is also weird to me. I haven't encountered this problem yet. I’m sorry that I do not have a good idea for you.

Best, Jun

YangHan-Morningstar commented 2 years ago

After that I continue to train until it's end and found it became converges after some epoch, I do think it's weird. But when I test the final model, I got the test result which was still little worse than the result in paper.

image image