search

YudeWang / SEAM

Self-supervised Equivariant Attention Mechanism for Weakly Supervised Semantic Segmentation, CVPR 2020 (Oral)
MIT License
539 stars 97 forks source link

CRF Inference #2

Closed won-bae closed 4 years ago

won-bae commented 4 years ago

Hi,

Thank you for sharing the code. When I was trying to understand the code, I ran into a trouble to understand crf_inference. In line 96 of infer_SEAM.py, bgcam_score does not seem like probabilities (I checked max value for some images are 1.2). But unary_from_softmax takes probs from softmax as inputs. I am not sure if something is wrong here or I am missing something. It would be highly appreciated if you could clarify it. Thank you.

YudeWang commented 4 years ago

Hi @won-bae , the crf inference code in infer_SEAM.py is prepared for the following AffinityNet training and generates foreground/background/undefined three parts. Here the bg_score is defined by parameter alpha and bgcam_score is not the probability in the strict sense. This part of code comes from the AffinityNet project. And you can combine it with VOC12AffDataset in /voc12/data.py to understand how it works.

won-bae commented 4 years ago

Thank you for the clarification!