A Taxonomy and Library for Visualizing Learned Features in Convolutional Neural Networks

TL;DR

This paper presents a comprehensive taxonomy for CNN visualization methods and introduces FeatureVis, an open-source library that facilitates understanding and analyzing learned features in convolutional neural networks.

Contribution

It provides a unified classification framework for CNN visualization techniques and offers an accessible, extendable library to aid researchers in analyzing learned features.

Findings

Classifies CNN visualization methods into three main categories.

Provides an open-source library with implementations for each class.

Enhances understanding of CNN intermediate layer features.

Abstract

Over the last decade, Convolutional Neural Networks (CNN) saw a tremendous surge in performance. However, understanding what a network has learned still proves to be a challenging task. To remedy this unsatisfactory situation, a number of groups have recently proposed different methods to visualize the learned models. In this work we suggest a general taxonomy to classify and compare these methods, subdividing the literature into three main categories and providing researchers with a terminology to base their works on. Furthermore, we introduce the FeatureVis library for MatConvNet: an extendable, easy to use open source library for visualizing CNNs. It contains implementations from each of the three main classes of visualization methods and serves as a useful tool for an enhanced understanding of the features learned by intermediate layers, as well as for the analysis of why a network might fail for certain examples.