search

linyq2117 / CLIP-ES

MIT License
175 stars 9 forks source link

Some details #17

Closed zbb1111 closed 6 months ago

zbb1111 commented 6 months ago

Sorry to bother you again, could you please tell me about the innovative points you proposed in which files and which part of the code? Like:

  1. Sharpness based prompt?
  2. Synonym fusion?
  3. CAA?
linyq2117 commented 6 months ago
  1. The prompt we used is a clean origami {}., which can be found at here.
  2. The synonyms of class names are shown in clip_text.py.
  3. The CAA is conducted at these lines.
zbb1111 commented 6 months ago

Thank you very much and wish you all the best!

lk0429 commented 6 months ago

Excuse me for disturbing, could you please tell me which part of the code implements the Confidence-guided Loss (CGL)?

linyq2117 commented 6 months ago

Excuse me for disturbing, could you please tell me which part of the code implements the Confidence-guided Loss (CGL)?

Thanks for your interest in our work!

We do not explicitly adopt confidence-guided loss. Instead, we set the segmentation mask to 255 if the max confidence of this pixel is less than 0.95 for convenience. The label 255 will be ignored by nn.CrossEntropyLoss in deeplab-pytorch. This process is implemented in eval_cam_with_crf.py.

linyq2117 commented 6 months ago

Closed for inactivity.