ANOMALY DETECTION USING DIFFUSION MODELS

Summary from the TLDR AI newsletter

DiAD leverages the power of diffusion models to detect anomalies. It uniquely combines a pixel-space autoencoder, a Semantic-Guided (SG) network, and a feature-space extractor to accurately detect and localize anomalies in multi-class settings.