https://github.com/deepankaracharyya/intermediate_representation

In this project, I will try to present a visual representation of the what the convnets learn during training.
https://github.com/deepankaracharyya/intermediate_representation

deep-neural-networks keras keras-models representation-learning

Last synced: 6 months ago
JSON representation

In this project, I will try to present a visual representation of the what the convnets learn during training.

• Host: GitHub
• URL: https://github.com/deepankaracharyya/intermediate_representation
• Owner: DeepankarAcharyya
• License: mit
• Created: 2020-03-24T10:36:10.000Z (almost 6 years ago)
• Default Branch: master
• Last Pushed: 2020-03-24T19:50:32.000Z (almost 6 years ago)
• Last Synced: 2025-05-07T06:51:38.599Z (8 months ago)
• Topics: deep-neural-networks, keras, keras-models, representation-learning
• Language: Jupyter Notebook
• Size: 5.58 MB
• Stars: 1
• Watchers: 1
• Forks: 0
• Open Issues: 0
• Metadata Files:
  • Readme: README.md
  • License: LICENSE

Awesome Lists containing this project

README

          
# Intermediate_Representation

In this project, I will try to present a visual representation of the what the convnets learn during training. 

The representations learned by convnets are highly amenable to visualization, in large because they are representations of visual concepts.

# Findings:

> * First Layer acts as a collection of various edge detectors.

> > At this stage, almost all the information present in the initial image is present.

> * As we go deeper, the activations becomes more abstract and less visually interpretable.

> > Higher presentations carry increasingly less information about the visual contents of the image and increasingly more information related to the class of the image.

> * In the following layers, more and more filters are blank, which means that the pattern encoded by the filter isn't found in the input image.