zero mean intensity of gradient for some cases about keras-grad-cam HOT 9 OPEN

I got the same problem and haven't solved it yet (¦3」∠)

from keras-grad-cam.

I did solve my error for the case that i did not choose the right layer output as the target.

from keras-grad-cam.

What layer did you take? Or what is the correct layer?

from keras-grad-cam.

@janphhe In my project, I used Xception. The target layer will be the global_avg_pooling layer, we should get a vector which seemed to be flattened. The grad visualization was made on it.

from keras-grad-cam.

Somebody has a solution for this. I get the same issue with a VGG19 network :/ I appreciate every help! Thanks a lot!

from keras-grad-cam.

I'm having the same problem, i'm sure I have selected the correct conv layer as it works for some images and not others. I'm not using a pre-trained model.

EDIT: Change to using Leaky-relu activations to prevent the vanishing gradient problem has solved it for me, all though this probably isn't easy with pre-trained networks

from keras-grad-cam.

@Ada-Nick, can you please give more details about your solution/intuition?

from keras-grad-cam.

@margiki Leaky-relu allows the gradient of negative values to be non-zero, preventing pooled_grads from being a null-matrix

from keras-grad-cam.

In my case, using LeakyReLU didn't solve the issue.

I delved deeper and found that the gradients computed by GradCAM were actually negative. Then, GradCAM applies a ReLU and the result was an empty map.

from keras-grad-cam.