MSEED features a hierarchical encoder-decoder architecture, a short-term-enhanced subnet, and a feature assembling layer that integrates spatial and temporal information across multivariate inputs.

This article investigates the potential of dual CNN-Transformer architectures for generalizable few-shot anomaly detection (GFSAD), a practical yet understudied form of anomaly detection (AD). In ...