search

dongzhang89 / CONTA

Implementation for paper: Causal Intervention for Weakly-Supervised Semantic Segmentation
MIT License
185 stars 27 forks source link

About M_t #12

Closed suyukun666 closed 3 years ago

suyukun666 commented 3 years ago

Nice job! Here is my question about the main operation for Eq(3). (1) In which py file do you operate for Eq(3)? (2) As I can see, in "segmentation/tools/m_t.py", where is the "cfg.DIC_FILE" and "cfg.PRIOR_PROB"?

dongzhang89 commented 3 years ago

@suyukun666 Thanks for your attention. 1. Eq(3) is performed in m_t.py. 2. "cfg.DIC_FILE" and "cfg.PRIOR_PROB" denotes the confounder set C and P(c), respectively. The confounder set can be obtained via computing the class-specific average segmentation-mask, and P(c) is set as the uniform 1/n. where n is the class number.

cyy53589 commented 3 years ago

@suyukun666 Thanks for your attention. 1. Eq(3) is performed in m_t.py. 2. "cfg.DIC_FILE" and "cfg.PRIOR_PROB" denotes the confounder set C and P(c), respectively. The confounder set can be obtained via computing the class-specific average segmentation-mask, and P(c) is set as the uniform 1/n. where n is the class number.

Hi, Thank you for your great work! You mentioned that Eq(3) is performed in m_t.py. I don't find any .py code importing m_t. Do I need run m_t.py manually ? @ZHANGDONG-NJUST

dongzhang89 commented 3 years ago

@cyy5358 m_t is used for the new round of classification. You can add it manually to make it work.

Welllee12366 commented 3 years ago

@cyy5358 m_t is used for the new round of classification. You can add it manually to make it work.

Sorry, as a beginner, I don't know which classifier this file needs to be added to. If it is convenient for you, I'm looking forward to your more detailed answer to this question. Thank you

dongzhang89 commented 3 years ago

@Welllee12366 If you don’t mind, you can send me an email at dongzhang@njust.edu.cn. I can give you my phone number, discussion on the phone may be more efficient