https://github.com/vermavinay982/feature-map-visualizer-grad-ascent

What a neural network captures? The filter and feature-maps can be visualized with any image to test with.
https://github.com/vermavinay982/feature-map-visualizer-grad-ascent

cnn deep deep-learning dl feature filter keras maps neural-network opencv python tensorflow

Last synced: 9 days ago
JSON representation

What a neural network captures? The filter and feature-maps can be visualized with any image to test with.

• Host: GitHub
• URL: https://github.com/vermavinay982/feature-map-visualizer-grad-ascent
• Owner: vermavinay982
• Created: 2021-01-25T18:23:11.000Z (almost 4 years ago)
• Default Branch: main
• Last Pushed: 2021-06-14T05:53:45.000Z (over 3 years ago)
• Last Synced: 2024-10-14T14:11:51.045Z (24 days ago)
• Topics: cnn, deep, deep-learning, dl, feature, filter, keras, maps, neural-network, opencv, python, tensorflow
• Language: Jupyter Notebook
• Homepage:
• Size: 38.2 MB
• Stars: 1
• Watchers: 1
• Forks: 0
• Open Issues: 0
• Metadata Files:
  • Readme: README.md

Awesome Lists containing this project

README

        
# Weights-Visualizer-for-CNN

What a Neural Network Captures? The filter and feature-maps can be visualized with any image to test with.

Feauture maps closer to the input of model captures a lot of fine detail in the image as we progress deeper in the model, feature map show less detail.

## Heatmaps - What makes Cat different from Monkey





 

 

 



 


## Convnets Filters - What Activates the ConvNet Most













## Feature Map on Bird Image - How Input Changes 

  

This pattern was expected as the model abstracts the features from the image into more general concepts that can be used to make a classification still it is not clear from final image the model saw a bird, we generally lose the ability to interpret these deeper feature maps