Semi-Supervised Bolt Anomaly Detection Based on Local Feature Reconstruction

Multimodal Anomaly Detection for High-Speed Train Control System Based on Attention Mechanism

“…The basic idea is to employ neural networks to reconstruct the anomalous regions, and detect anomalies by calculating the error between the original and reconstructed images (Jiang et al 2023). Due to the uncertainties of the image environment such as noise interference, various lighting conditions, and variable camera viewing angles (Peng et al 2023), image anomaly detection faces numerous challenges, resulting in low detection performance. In response to these, researchers from China and other countries have proposed various new solutions.…”

“…In response to these, researchers from China and other countries have proposed various new solutions. For instance, Peng et al (2023) proposed a novel semi-supervised anomaly detection algorithm that learns from both anomaly-free samples and annotations, while being able to identify anomaly-free and anomalous targets. (2023) proposed a mechanical anomaly detection method for complex working conditions -a full graph dynamic autoencoder.…”