keravis 1.0.0

A high-level API for ConvNet visualizations in Keras

# keravis

keravis is a high-level API for ConvNet visualizations in Keras. As of v1.0, it supports visualizations of

1. Convolutional layer activations

2. 2-dimensional feature space representations

3. Saliency maps (vanilla backprop, guided backprop, and occlusion)

4. Synthetic maximally-activating images of classifier output

5. Maximally activating patches of an intermediate neuron in a set of images

with support for nested pretrained models.

This is a hobby project that was inspired by lecture 14 of Stanford’s CS231n: Convolutional Neural Networks for Visual Recognition http://cs231n.stanford.edu/. It is not yet optimized for serious use (see keras-vis instead).

## Installation ## Usage

## MNIST Examples `python from keravis import feature_space feature_space(model,X=x_test[:5000],y=y_test[:5000],kind='tsne') ` 

`python from keravis import saliency_backprop saliency_backprop(model,test_img,class_idx=7) ` 

`python from keravis import saliency_guided_backprop saliency_guided_backprop(model,test_img,class_idx=7) ` 

`python from keravis import classifier_gradient_ascent classifier_gradient_ascent(model,class_idx=5,dim=(28,28,1)) ` 

`python from keravis import maximally_activating_conv_features maximally_activating_conv_features(model,'conv2d_1',X=x_test) `