Software & Data Downloads — LTAD

Long-Tailed Anomaly Detection (LTAD) Dataset for Anomaly detection (AD) performance evaluation.

Anomaly detection (AD) aims to identify defective images and localize their defects (if any). Ideally, AD models should be able to: detect defects over many image classes; not rely on hard-coded class names that can be uninformative or inconsistent across datasets; learn without anomaly supervision; and be robust to the long-tailed distributions of real-world applications. To address these challenges, we formulate the problem of long-tailed AD by introducing several datasets with different levels of class imbalance for performance evaluation.

  •  Ho, C.-H., Peng, K.-C., Vasconcelos, N., "Long-Tailed Anomaly Detection with Learnable Class Names", IEEE Conference on Computer Vision and Pattern Recognition (CVPR), Farhadi, A. and Crandall, D. and Sato, I. and Wu, J. and Pless, R. and Akata, Z., Eds., DOI: 10.1109/​CVPR52733.2024.01182, June 2024, pp. 12435-12446.
    BibTeX TR2024-040 PDF Video Data Presentation
    • @inproceedings{Ho2024jun,
    • author = {Ho, Chih-Hui and Peng, Kuan-Chuan and Vasconcelos, Nuno},
    • title = {Long-Tailed Anomaly Detection with Learnable Class Names},
    • booktitle = {IEEE Conference on Computer Vision and Pattern Recognition (CVPR)},
    • year = 2024,
    • editor = {Farhadi, A. and Crandall, D. and Sato, I. and Wu, J. and Pless, R. and Akata, Z.},
    • pages = {12435--12446},
    • month = jun,
    • publisher = {IEEE},
    • doi = {10.1109/CVPR52733.2024.01182},
    • issn = {2575-7075},
    • isbn = {979-8-3503-5300-6},
    • url = {https://www.merl.com/publications/TR2024-040}
    • }

Access data at https://doi.org/10.5281/zenodo.10854201.