This repository is a curated list of papers, datasets, and recent research in industrial image anomaly/defect detection. It serves as a comprehensive resource for researchers and practitioners in the field, aiming to provide an organized overview of state-of-the-art methods and available benchmarks.

How It Works

The repository categorizes research by anomaly detection approach (e.g., feature-embedding, reconstruction-based, supervised), research direction (e.g., zero-shot, noisy AD, anomaly synthesis), and specific conference venues. It also includes a detailed table of datasets with their characteristics and links to related papers and code.

Quick Start & Requirements

This repository is a collection of research papers and datasets, not a runnable software library. To utilize the resources, users would need to access the linked papers and datasets independently.

Highlighted Details

• Extensive categorization of anomaly detection methods, including emerging areas like Vision-Language models and Diffusion Models.
• Comprehensive listing of over 30 industrial anomaly detection datasets with detailed statistics.
• Regular updates on recent research from top computer vision conferences (CVPR, ECCV, NeurIPS, etc.).
• Links to associated code repositories for many listed papers.

Maintenance & Community

The repository is actively maintained and welcomes contributions for categorizing papers and adding new resources. It links to a survey paper and a benchmark paper, indicating ongoing research engagement.

Licensing & Compatibility

The repository itself is likely under a permissive license (e.g., MIT, Apache), but the licensing of the linked papers and datasets would vary and must be checked individually.

Limitations & Caveats

This is a curated list and does not provide a unified framework or software for performing anomaly detection. Users must independently find, download, and implement the methods and datasets.