search

utkuozbulak / pytorch-cnn-visualizations

Pytorch implementation of convolutional neural network visualization techniques
MIT License
7.85k stars 1.49k forks source link

Sharp edges in layer visualization #22

Closed McLawrence closed 6 years ago

McLawrence commented 6 years ago

This is probably not an issue with your code as it does not happen for your examples:

When doing activation maximization, the right and bottom rim of the generated image has a sharp edge, were the image grey: test_edge

Do you have any idea why this jump occurs? Is this a problem with the model I am using or with the activation maximization itself?

utkuozbulak commented 6 years ago

Hey,

I vaguely remember coming across the same type of results but I can't recall why and/or how I fixed it (or even if I did fix it). Sorry for not being of help on this issue.