How convolutional neural network see the world - A survey of convolutional neural network visualization methods

TL;DR

This survey reviews various CNN visualization techniques that help interpret the internal features of convolutional neural networks, enhancing understanding and practical application in computer vision tasks.

Contribution

It provides a comprehensive overview of key CNN visualization methods, detailing their motivations, algorithms, and experimental results, and discusses their practical applications.

Findings

Different visualization methods reveal diverse internal features.

Visualization techniques improve CNN interpretability and trust.

Applications include network design, optimization, and security.

Abstract

Nowadays, the Convolutional Neural Networks (CNNs) have achieved impressive performance on many computer vision related tasks, such as object detection, image recognition, image retrieval, etc. These achievements benefit from the CNNs outstanding capability to learn the input features with deep layers of neuron structures and iterative training process. However, these learned features are hard to identify and interpret from a human vision perspective, causing a lack of understanding of the CNNs internal working mechanism. To improve the CNN interpretability, the CNN visualization is well utilized as a qualitative analysis method, which translates the internal features into visually perceptible patterns. And many CNN visualization works have been proposed in the literature to interpret the CNN in perspectives of network structure, operation, and semantic concept. In this paper, we expect to provide a comprehensive survey of several representative CNN visualization methods, including Activation Maximization, Network Inversion, Deconvolutional Neural Networks (DeconvNet), and Network Dissection based visualization. These methods are presented in terms of motivations, algorithms, and experiment results. Based on these visualization methods, we also discuss their practical applications to demonstrate the significance of the CNN interpretability in areas of network design, optimization, security enhancement, etc.