search

caoyunkang / Segment-Any-Anomaly

Official implementation of "Segment Any Anomaly without Training via Hybrid Prompt Regularization (SAA+)".
704 stars 76 forks source link

About classification #12

Closed laiyingxin2 closed 10 months ago

laiyingxin2 commented 1 year ago

Thank you so much for your wonderful paper! I have really enjoyed it! I have a question regarding the SSA,

  1. What is the model's classification result for abnormal/non-abnormal in the entire picture?
  2. What is the cross-domain effect of classification?
  3. I noticed that adding a classification loss can lead to overfitting and non-convergence during fine-tuning. Could you please share any approaches you have used to address this problem? Thank you very much for your time and expertise. I greatly appreciate any insights or suggestions
caoyunkang commented 1 year ago

Hi! Thanks for your interest. Much to our regret, SAA's classification performance is subpar as it always selects the most significant regions to be anomalies, i.e., it tends to be false-alarmed. You could directly run our code to evaluate the classification performance. Best :)