Skip to content

himanshurawlani/convnet-interpretability-keras

Repository files navigation

Visual intrepretability for Convolutional Neural Network in Keras

Following visualization techniques are used:

  • Visualizing intermediate activations (convolution outputs)
  • Visualizing convolutional filter/kernels
  • Visualizing input pixel space from intermediate activation using deconvnets
  • Visualizing areas of interest in input image for each layer using GradCAM technique

Read more about this project in my blog here