Anomaly Detection with Adversarial Dual Autoencoders

TL;DR

This paper introduces ADAE, a GAN-based anomaly detection framework using dual autoencoders to improve training stability and detection accuracy, demonstrated effectively on diverse datasets including brain tumor detection.

Contribution

The paper proposes a novel GAN-based anomaly detection method with dual autoencoders, enhancing training stability and detection performance over existing approaches.

Findings

Demonstrates robustness across multiple datasets

Effective in brain tumor detection scenarios

Improves training stability of GAN-based models

Abstract

Semi-supervised and unsupervised Generative Adversarial Networks (GAN)-based methods have been gaining popularity in anomaly detection task recently. However, GAN training is somewhat challenging and unstable. Inspired from previous work in GAN-based image generation, we introduce a GAN-based anomaly detection framework - Adversarial Dual Autoencoders (ADAE) - consists of two autoencoders as generator and discriminator to increase training stability. We also employ discriminator reconstruction error as anomaly score for better detection performance. Experiments across different datasets of varying complexity show strong evidence of a robust model that can be used in different scenarios, one of which is brain tumor detection.