I'm new to deep learning, so maybe this is a silly question...

Do any adjustments need to be made for applying Grad-CAM on CNNs that use a Global Average Pooling (GAP) layer right before fully connected ones?

I understand that the GAP layer aggregates the activations of an intermediate layer in order to produce a compact representation of the image, removing information regarding the features location. Is this an obstacle to grad-cam backpropagation?

I imagine that for a CNN that uses, for example, a Max Pooling layer followed by a Flatten layer, o Grad-CAM is capable of retriving the exact location of the relevant features.

I'm sorry if it is a silly doubt, but I couldn't find the answer for it anywhere.

Thanks in advance!

  • $\begingroup$ Why do you think a global max pooling layer would retrieve the exact location of relevant features? Also, max pooling layers could be made differentiable so backpropagation isn't a problem $\endgroup$ Commented Feb 6, 2023 at 15:37
  • $\begingroup$ Actually, I was trying to say that I don't think that GAP can retrieve the feature location. Thus, would it be necessary to make some kind adjustment to apply Grad-CAM? I'm asking that because I have the impression that Grad-CAM, sometimes, highlights areas which are larger than necessary. I thought the GAP could be a reason Also, thanks for the "max pooling link", really interesting! $\endgroup$ Commented Feb 8, 2023 at 0:01


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Browse other questions tagged or ask your own question.