search

Megvii-BaseDetection / DisAlign

Implementation of "Distribution Alignment: A Unified Framework for Long-tail Visual Recognition"(CVPR 2021)
Apache License 2.0
118 stars 10 forks source link

about the confidence score σ(x) #9

Closed lzed2399 closed 3 years ago

lzed2399 commented 3 years ago

In the paper, the σ(x) is implemented as a linear layer followed by a non-linear activation function (e.g., sigmoid function) for all input x. How to understand the input x？the matrix of raw iamge, or the extracted features, even or cls_score? Thank you!

tonysy commented 3 years ago

Hi, the x of \sigma(x) is the extracted features.

tonysy commented 3 years ago

Feel free to reopen this issue if you have any questions.