Hi, I implemented the Visualization of the attention module part in the paper by using a row of energy matrix in your codes and the given pre-trained model pam, the result seem not well, especially the pixels belonging to building are less attentive to all the pixels and least attentive to pixels belonging to building. Maybe It‘s because I used the wrong part of the attention module for implementation. Therefore, Would you mind to tell which part of the attention module is used to implement the visualization of the attention module in your paper? Many thanks.
Hi, I implemented the Visualization of the attention module part in the paper by using a row of energy matrix in your codes and the given pre-trained model pam, the result seem not well, especially the pixels belonging to building are less attentive to all the pixels and least attentive to pixels belonging to building. Maybe It‘s because I used the wrong part of the attention module for implementation. Therefore, Would you mind to tell which part of the attention module is used to implement the visualization of the attention module in your paper? Many thanks.