understand_cnn

Reproduce the paper "Visualizing and Understanding Convolutional Networks"

And the implementation is transformed from a Kera based repo.

https://github.com/saketd403/Visualizing-and-Understanding-Convolutional-neural-networks

By run the following command you should be able to get the figures.

python main.py

And here is an example image to visualize the 4th layer in the network.

Issues:

1. The bias term of convolution is never used.
2. The linear layer is hard to be visualized.
3. Start from 16th layer, the visualization is becoming harder and harder.