How to Visualize Attention Mechanism

[pfeducode]

Is there any code for visualizing the attention mechanism? Or can you recommend a repository for reference，Thank you

[harlanhong]

I wrote it by myself. You can refer to the following code:

[pfeducode]

This is my modified code. The result is as follows, but my result seems different from yours. Is my result correct?

    def visualize_attention2(self, source_image,attention):
        amap=attention.detach().cpu().numpy()

        img=source_image.detach().cpu().numpy().copy()
        img=np.squeeze(img)
        img=img.transpose((1,2,0))

        mask=amap.reshape(1,4096,64,64)
        mask = np.mean(mask, axis=1) #[1,64,64]
        mask = np.squeeze(mask, axis=0) #[64,64]
        mask=(mask-mask.min())/(mask.max()-mask.min())
        #mask=mask/mask.max()
        mask=cv2.resize(mask,(img.shape[1],img.shape[0]))
        img=np.float32(img)/255
        heatmap=cv2.applyColorMap(np.uint8(255*mask),cv2.COLORMAP_JET)
        heatmap=np.float32(heatmap)/255
        cam=heatmap+np.float32(img)
        cam=cam/np.max(cam)
        img=np.uint8(255*cam)

        return img

if i use opencv-python to save image,the result as follows if i use imageio.imsave to save image,the result as follows As the number of iterations increases, the face disappears

[harlanhong]

you should make sure that your input "source_image" ranges between 0 to 255. It seems that you put these codes in the generator.py, and the source_image in the generator.py is normalised to [0,1]. You should check that.

[pfeducode]

I have got the correct picture, thank you for your reply and guidance, and thank you again for your excellent work.