https://github.com/abhijit-2592/visualizing_cnns

An attempt to visualize Convolutional Neural Networks
https://github.com/abhijit-2592/visualizing_cnns

Last synced: 7 days ago
JSON representation

An attempt to visualize Convolutional Neural Networks

• Host: GitHub
• URL: https://github.com/abhijit-2592/visualizing_cnns
• Owner: Abhijit-2592
• License: mit
• Created: 2020-03-05T02:31:52.000Z (almost 5 years ago)
• Default Branch: master
• Last Pushed: 2020-03-05T19:54:46.000Z (almost 5 years ago)
• Last Synced: 2024-12-03T11:12:45.122Z (about 1 month ago)
• Language: Jupyter Notebook
• Homepage:
• Size: 9.51 MB
• Stars: 0
• Watchers: 0
• Forks: 0
• Open Issues: 0
• Metadata Files:
  • Readme: README.md
  • License: LICENSE

Awesome Lists containing this project

README

        
# Visualizing and understanding Convolutional Neural Networks

This repository is an attempt to code up a few visualization techniques in **Tensorflow 2** mentioned by [Andrej Karpathy](https://twitter.com/karpathy) in his [Lecture](https://www.youtube.com/watch?v=ta5fdaqDT3M&list=PLkt2uSq6rBVctENoVBg1TpCC7OQi31AlC&index=9) ([Slides](http://cs231n.stanford.edu/slides/2016/winter1516_lecture9.pdf)). The notebooks will directly run on **Google Colab**. You just need to upload the example images present in the [images](./images) folder to **Colab** while running these notebooks. I have also given links to the corresponding papers and sources within each notebook.

It is recommended that you see the notebooks in the order given below as it builds up from intuition based visualization techniques to optimization based techniques.

1.  occlusion_experiment.ipynb

2.  filter_visualization.ipynb

3.  activation_maximization.ipynb

4.  saliency_map.ipynb

5.  grad_cam.ipynb

I have tried my best to reproduce the techniques presented in those papers. Please feel free to provide any improvements and suggestions.