CNN-Visualization - GitLite

KO | EN

gitlite — search

#javascript #python #llm #go #android #deep-learning #linux #docker #macos #nodejs #vue #awesome-list #ios #openai

CNN-Visualization

(★ 197)

TensorFlow implementations of visualization of convolutional neural networks, such as Grad-Class Activation Mapping and guided back propagation

File Explorer

Download Latest Version (.zip)

- cat.jpg
- im_0.png
- test_0.png
- - README.md
- - README.md
- - README.md
- - README.md
- - README.md
- cam.py
- config_cam.py
- config_path.py
- deconv.py
- gap.py
- gradcam.py
- guided_backpropagation.py
- invert.py
- run.sh
- setup_env.py
- vizfilter.py
- - __init__.py
  - cifar.py
  - image.py
- - __init__.py
  - cam.py
  - gap.py
  - grad_cam.py
  - guided_backpro.py
  - invert.py
- - __init__.py
  - googlenet.py
  - layers.py
  - vgg.py
- - __init__.py
  - image.py
  - normalize.py
  - viz.py
- __init__.py
- setup_test_env.py
- test_cam.py
- test_gradcam.py
- test_guided_backprop.py
.gitignore
.travis.yml
LICENSE
README.md
requirements.txt

# Installation Guide

1. Get the code

git clone https://github.com/conan7882/CNN-Visualization

Downloads the entire project code from GitHub to your computer.

cd CNN-Visualization

Moves into the project folder you just downloaded.

2. Python

Easy Recommended

Prerequisites

Git Needed to download the project code from GitHub.
Python 3 On Windows, be sure to check 'Add Python to PATH' during installation.

pip install -r requirements.txt

Installs the Python libraries listed in requirements.txt (or similar).

python <실행할 파일명>.py # README에서 정확한 실행 파일명을 확인하세요

Runs the Python script (or module).

✅ If it runs without errors and prints output in the terminal, it worked.

// repository documentation

Was this content helpful?

★ 0 (0 ratings)

# Visualization of Deep Covolutional Neural Networks

- This repository contains implementations of visualizatin of CNN in recent papers.
- The source code in the repository can be used to demostrate the algorithms as well as test on your own data.

## Requirements
- Python 3.3+
- [Tensorflow 1.3](https://www.tensorflow.org/)
- [TensorCV](https://github.com/conan7882/DeepVision-tensorflow)

## Algorithms

- [Visulization of filters and feature maps of GoogLeNet](https://github.com/conan7882/CNN-Visualization/tree/master/doc/firstfilter#visualization-of-filters-and-feature-maps-of-googlenet)
- [Deconvolutional Networks](https://github.com/conan7882/CNN-Visualization/blob/master/doc/deconv/README.md#cnn-feature-visuallization-via-deconvnet-transposed-convolutional-layers) (ECCV'14)
- [Guided back propagation](https://github.com/conan7882/CNN-Visualization/tree/master/doc/guided_backpropagation#guided-backpropagation) (2014)
- [Class Activation Mapping (CAM)](https://github.com/conan7882/CNN-Visualization/tree/master/doc/cam#class-activation-mapping-cam) (CVPR'16)
- [Gradient-weighted Class Activation Mapping (Grad-CAM)](https://github.com/conan7882/CNN-Visualization/tree/master/doc/grad_cam#gradient-weighted-class-activation-mapping-grad-cam) (ICCV'17)

## [Visulization of filters and feature maps of GoogLeNet](https://github.com/conan7882/CNN-Visualization/tree/master/doc/firstfilter#visualization-of-filters-and-feature-maps-of-googlenet) 
- The most straightforward approach to visualize a CNN is to show the feature maps (activations) and filters.
- Details of the implementation and more results can be found [here](https://github.com/conan7882/CNN-Visualization/tree/master/doc/firstfilter#visualization-of-filters-and-feature-maps-of-googlenet)
<p align = 'left'>
<img src ="doc/firstfilter/figs/GoogLeNet.png" height="250" />
<img src ="doc/firstfilter/figs/GoogLeNet_inception3a.png" height="250" />
</p>

## [Deconvnet](https://github.com/conan7882/CNN-Visualization/blob/master/doc/deconv/README.md#cnn-feature-visuallization-via-deconvnet-transposed-convolutional-layers)
- Pick a specific activation on a feature map and set other activation to zeros, then reconstruct an image by mapping back this new feature map to input pixel space.
- Details of the implementation and more results can be found [here](https://github.com/conan7882/CNN-Visualization/blob/master/doc/deconv/README.md#cnn-feature-visuallization-via-deconvnet-transposed-convolutional-layers). Some results:

<p align = 'left'>
<img src ="doc/deconv/figs/people/conv1_2_feat.png" height="200" />
<img src ="doc/deconv/figs/people/conv1_2_im.png" height="200" />
<img src ="doc/deconv/figs/people/conv2_2_feat.png" height="200" />
<img src ="doc/deconv/figs/people/conv2_2_im.png" height="200" />
<img src ="doc/deconv/figs/people/conv3_4_feat.png" height="200" />
<img src ="doc/deconv/figs/people/conv3_4_im.png" height="200" />
<img src ="doc/deconv/figs/people/conv4_4_feat.png" height="200" />
<img src ="doc/deconv/figs/people/conv4_4_im.png" height="200" />
</p>

## [Guided back propagation](https://github.com/conan7882/CNN-Visualization/tree/master/doc/guided_backpropagation#guided-backpropagation)

- Details of the implementation and more results can be found [here](https://github.com/conan7882/CNN-Visualization/tree/master/doc/guided_backpropagation#guided-backpropagation). Some results:

![gbp](https://raw.githubusercontent.com/conan7882/CNN-Visualization/master/doc/guided_backpropagation/figs/gbp.png)

## [Class Activation Mapping (CAM)](https://github.com/conan7882/CNN-Visualization/tree/master/doc/cam#class-activation-mapping-cam)
- The class activation map highlights the most informative image regions relevant to the predicted class. This map can be obtained by adding a global average pooling layer at the end of convolutional layers.
- Details of the implementation and more results can be found [here](https://github.com/conan7882/CNN-Visualization/tree/master/doc/cam#class-activation-mapping-cam). Some results:

![celtech_change](https://raw.githubusercontent.com/conan7882/CNN-Visualization/master/doc/cam/figs/celtech_diff.png)

## [Gradient-weighted Class Activation Mapping (Grad-CAM)](https://github.com/conan7882/CNN-Visualization/tree/master/doc/grad_cam#gradient-weighted-class-activation-mapping-grad-cam)
- Grad-CAM generates similar class heatmap as CAM, but it does not require to re-train the model for visualizatin.
- Details of the implementation and more results can be found [here](https://github.com/conan7882/CNN-Visualization/tree/master/doc/grad_cam#gradient-weighted-class-activation-mapping-grad-cam). Some results:

![grad-cam-result](https://raw.githubusercontent.com/conan7882/CNN-Visualization/master/doc/grad_cam/figs/ex1.png)