Is it correct to freeze the weight and bias of the DisAlign Linear Layer as well?

[jeongHwarr]

Hello. Thank you for your project! I'm testing your code on my custom dataset. My task is classification. I have a question about your code implementation.

https://github.com/Megvii-BaseDetection/DisAlign/blob/a2fc3500a108cb83e3942293a5675c97ab3a2c6e/classification/imagenetlt/resnext50/resx50.scratch.imagenet_lt.224size.90e.disalign.10e/net.py#L56-L62

From my understanding, in stage 2, remove the linear layer used in stage 1 and add DisAlign Linear Layer. And freeze all parts except for logit_scale, logit_bias, and confidence_layer. At this time, the weight and bias of DisAlignLinear are also frozen. (self.weight, self.bias) Is my understanding correct?

If so, are the weight and bias of DisAlignLinearLayer fixed after the initialization? (The weight and bias of the linear layer in stage 1 are not copied either)

If my understanding is correct, why is the weight of DisAlignLinear also frozen?

I will wait for your reply. thanks!

[tonysy]

Hi, in our proposed method, we keep the original classifier and add scale/bias for stage-2 learning. Please read the adaptive calibration function section of the original paper.

[jeongHwarr]

@tonysy Thank you for your reply! I thought that the original classifier would be removed and replaced with a new classifier.

https://github.com/Megvii-BaseDetection/DisAlign/blob/a2fc3500a108cb83e3942293a5675c97ab3a2c6e/classification/imagenetlt/resnext50/resx50.scratch.imagenet_lt.224size.90e.disalign.10e/net.py#L21-L27

I couldn't run the cvpod because my os is Windows.

Does the above code means a classifier in which other elements are added while retaining the original classifier weight?

[tonysy]

Does the above code means a classifier in which other elements are added while retaining the original classifier weight?

Yes

[jeongHwarr]

@tonysy Okay. Thank you!