search

torrvision / crfasrnn

This repository contains the source code for the semantic image segmentation method described in the ICCV 2015 paper: Conditional Random Fields as Recurrent Neural Networks. http://crfasrnn.torr.vision/
Other
1.34k stars 462 forks source link

does not need backward computation #106

Open Dan1900 opened 7 years ago

Dan1900 commented 7 years ago

when i train my own model, i found all layers do not need backward computation,but there is nothing error information,why?

nathanin commented 7 years ago

@Dan1900 Have you solved this? Was it an error with your build, or something to be set in a prototxt file?

bittnt commented 7 years ago

Did you set up the flag to force to not backward in prototxt?

nathanin commented 7 years ago

I've since solved this. In my prototxt I had loss_weight: 1 in a custom softmax loss layer.. removing that seems to have fixed it. On the other hand it could be something else but the broader point is it works now :)

MeisonP commented 6 years ago

@Dan1900 hello, have u fixed it? I got the same problem......

Duke-ding commented 5 years ago

i also got the same problem? how to deal with this issue.

SJie1997 commented 5 years ago

Has anyone solved this problem? I also encountered this problem and did not find obvious errors.