Awesome Open Source
Awesome Open Source

convisualize

Visualizations for Convolutional Neural Networks (CNNs) in Pytorch

The corresponding article can be found here!

Requirements:

  • Pytorch
  • Torchvision
  • Numpy
  • Matplotlib
  • Pillow

Note: In case you don't have a GPU, remove all instances of "cuda" and "cpu" from the notebook before running.

TODO

  • [X] Layer Outputs at all layers
  • [X] Filter outputs at a given layer
  • [X] Filter visualization at a given layer
  • [X] Image heatmap using Occlusion
  • [ ] Image heatmap using Grad Cam
  • [X] Class specific saliency maps
  • [X] SmoothGrad
  • [ ] Semantic segmentation using GrabCut
  • [X] Visualization of class models (Gradient Ascent)
  • [ ] Regularization techniques for class models (L2, Clip, Blur, etc.)
  • [ ] Guided Backprop
  • [ ] Filter visualization (Gradient Ascent)
  • [ ] Neural Texture Synthesis
  • [ ] Deep Dream
  • [ ] Neural Style Transfer

References



Alternative Project Comparisons
Related Awesome Lists
Top Programming Languages
Top Projects

Get A Weekly Email With Trending Projects For These Topics
No Spam. Unsubscribe easily at any time.
Python (840,414
Jupyter Notebook (163,805
Deep Learning (37,771
Network (37,571
Pytorch (21,715
Neural (16,616
Neural Network (15,797
Visualization (15,440
Convolutional Neural Networks (12,650
Cs231n (311
Texture Synthesis (141