Advances in Anomaly Detection and Synthesis with Diffusion Models

The field of anomaly detection and synthesis is rapidly advancing with the development of innovative diffusion-based models. These models have shown significant promise in improving the accuracy and efficiency of anomaly detection and synthesis tasks. A key direction in this field is the integration of diffusion models with other techniques, such as denoising autoencoders and conditional diffusion, to enhance their performance. Another important trend is the development of robust and efficient diffusion models that can handle complex and diverse data distributions. Noteworthy papers in this area include: Diffusion-Scheduled Denoising Autoencoders for Anomaly Detection in Tabular Data, which proposes a framework that integrates diffusion-based noise scheduling and contrastive learning into the encoding process. RDDPM: Robust Denoising Diffusion Probabilistic Model for Unsupervised Anomaly Segmentation, which offers flexibility in constructing various robust diffusion models. Single-Step Reconstruction-Free Anomaly Detection and Segmentation via Diffusion Models, which introduces a reconstruction-free approach that directly produces anomaly maps from the diffusion model, improving both detection accuracy and computational efficiency.

Sources

Diffusion-Scheduled Denoising Autoencoders for Anomaly Detection in Tabular Data

RDDPM: Robust Denoising Diffusion Probabilistic Model for Unsupervised Anomaly Segmentation

MissDDIM: Deterministic and Efficient Conditional Diffusion for Tabular Data Imputation

SARD: Segmentation-Aware Anomaly Synthesis via Region-Constrained Diffusion with Discriminative Mask Guidance

Quality-Aware Language-Conditioned Local Auto-Regressive Anomaly Synthesis and Detection

Single-Step Reconstruction-Free Anomaly Detection and Segmentation via Diffusion Models

Built with on top of