Meta-learning with GANs for anomaly detection, with deployment in high-speed rail inspection system

TL;DR

This paper introduces a meta-learning framework combined with GANs for anomaly detection in high-speed rail inspection, effectively handling scarce abnormal data and complex backgrounds, leading to significant workload reduction and efficiency improvements.

Contribution

The paper presents a novel meta-learning approach integrated with GANs and SSIM for robust anomaly detection in noisy, imbalanced datasets, specifically applied to high-speed rail inspection.

Findings

Achieved over 99.7% workload reduction in rail inspections

Saved 96.7% inspection time in real-world deployment

Demonstrated robustness with limited labeled data

Abstract

Anomaly detection has been an active research area with a wide range of potential applications. Key challenges for anomaly detection in the AI era with big data include lack of prior knowledge of potential anomaly types, highly complex and noisy background in input data, scarce abnormal samples, and imbalanced training dataset. In this work, we propose a meta-learning framework for anomaly detection to deal with these issues. Within this framework, we incorporate the idea of generative adversarial networks (GANs) with appropriate choices of loss functions including structural similarity index measure (SSIM). Experiments with limited labeled data for high-speed rail inspection demonstrate that our meta-learning framework is sharp and robust in identifying anomalies. Our framework has been deployed in five high-speed railways of China since 2021: it has reduced more than 99.7% workload and saved 96.7% inspection time.