search

suhwan-cho / TMO

[WACV 2023] Treating Motion as Option to Reduce Motion Dependency in Unsupervised Video Object Segmentation
MIT License
56 stars 3 forks source link

Generating single channel output #15

Closed JawadTawhidi closed 10 months ago

JawadTawhidi commented 10 months ago

Hi, is it possible to generate single channel output by TMO? as TMO currently uses standard cross entropy loss, you are generating 2 channel output, can we just change the last layer of TMO to generate single channel output? if so, then do we need to change the code of IOU after loss and also the code for J and F values inside mode=val ?

suhwan-cho commented 10 months ago

Hi,

TMO outputs 2-channel scores during training, and those scores are converted to 1-channel output using an argmax operation during testing.

JawadTawhidi commented 10 months ago

Thank you so much for your response. TMO is using standard cross entropy loss which accepts 2-channel scores, but I want to try TMO with binary cross entropy loss which expects single channel inputs. In that case can I just leave TMO as it is to generate 2-channel output, but before passing predicted scores to binary cross entropy loss change them to 1-channel output? similar the approach you have used for testing.

suhwan-cho commented 10 months ago

Yes, exactly!