Anomaly Detection by One Class Latent Regularized Networks

TL;DR

This paper introduces a novel adversarial dual autoencoder network for anomaly detection that improves training stability and accuracy by capturing data structure in latent space and using an auxiliary autoencoder as a discriminator.

Contribution

The paper proposes a new adversarial dual autoencoder architecture that enhances anomaly detection performance and training stability over existing GAN-based methods.

Findings

Achieves state-of-the-art results on MNIST and CIFAR10 datasets.

Demonstrates improved training stability with the auxiliary autoencoder.

Outperforms previous methods on GTSRB stop signs dataset.

Abstract

Anomaly detection is a fundamental problem in computer vision area with many real-world applications. Given a wide range of images belonging to the normal class, emerging from some distribution, the objective of this task is to construct the model to detect out-of-distribution images belonging to abnormal instances. Semi-supervised Generative Adversarial Networks (GAN)-based methods have been gaining popularity in anomaly detection task recently. However, the training process of GAN is still unstable and challenging. To solve these issues, a novel adversarial dual autoencoder network is proposed, in which the underlying structure of training data is not only captured in latent feature space, but also can be further restricted in the space of latent representation in a discriminant manner, leading to a more accurate detector. In addition, the auxiliary autoencoder regarded as a discriminator could obtain an more stable training process. Experiments show that our model achieves the state-of-the-art results on MNIST and CIFAR10 datasets as well as GTSRB stop signs dataset.