search

shjo-april / PuzzleCAM

[ICIP 2021] Puzzle-CAM: Improved localization via matching partial and full features.
174 stars 46 forks source link

Ask for details of the training process！ #15

Open YuYue26 opened 2 years ago

YuYue26 commented 2 years ago

I am trying to train with ResNest101, and I also added affinity and RW. When I try to train, it runs according to the specified code. It is found that the obtained affinity labels are not effective, and the effect of pseudo_labels is almost invisible, which is close to the effect of all black. I don't know where the problem is, who can explain the details. help!

qiang92 commented 2 years ago

I guess for the question "pseudo_labels is almost invisible", the reason maybe it generates binary segmentation masks which is a gary image. It is hard to see pixels in black background. You can use its trained model to get pseudo_labels to check if there is something wrong.