Inverting Visual Representations with Convolutional Networks

TL;DR

This paper introduces a novel method to invert and visualize visual feature representations from both shallow and deep neural networks, revealing rich information and interpretability of learned features.

Contribution

It presents a new up-convolutional network approach for inverting image representations, outperforming existing methods especially for shallow features, and offers insights into deep network representations.

Findings

Shallow features contain surprisingly rich information.

Deep network activations can reconstruct colors and contours.

Higher layers encode class-specific information.

Abstract

Feature representations, both hand-designed and learned ones, are often hard to analyze and interpret, even when they are extracted from visual data. We propose a new approach to study image representations by inverting them with an up-convolutional neural network. We apply the method to shallow representations (HOG, SIFT, LBP), as well as to deep networks. For shallow representations our approach provides significantly better reconstructions than existing methods, revealing that there is surprisingly rich information contained in these features. Inverting a deep network trained on ImageNet provides several insights into the properties of the feature representation learned by the network. Most strikingly, the colors and the rough contours of an image can be reconstructed from activations in higher network layers and even from the predicted class probabilities.