Pytorch-vis is a a neural network visualization toolkit for pytorch, which aims to provide easy and effective ways to visualize the trained models in pytorch. Pytorch-vis can be used seamlessly with pytorch, so you can visualize and have a deep insight into the trained model without pain. Please go to documentation for details.

Now pytorch-vis is still in development and will support following visualization techniques:

• Silency maps
• Class activation maps
• SmoothGrad
• Activation maximization
• ...

pytorch-vis is inspired by following greatful projects:

• keras-vis https://raghakot.github.io/keras-vis
• quiver https://keplr-io.github.io/quiver/
• pytorch-cnn-visualization https://github.com/utkuozbulak/pytorch-cnn-visualizations Thanks to the authors of these projects.

This project is under MIT LICENSE