You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Hi, congrats for the accept to CVPR2023!
I have two questions about the details after reading the paper.
Why use Hadamard product to get the ensemble of intermediate features and high-level features? Can we just use the average of the two? In addition, can AdaptiveAvgPool2d be used for downsampling for the feature alignment? What is its performance?
In the Preliminaries, you give an example of the shared layers of a ResNet model. I am curious about what exactly the S stage is in ResNet32. Are layer1 and layer2 the shared layers and layer 3 the exclusive layer?
The text was updated successfully, but these errors were encountered:
Hi, congrats for the accept to CVPR2023!
I have two questions about the details after reading the paper.
The text was updated successfully, but these errors were encountered: