3 predicting masked information

predicting masked information (e.g. patches, future frames, etc.) and leveraging the reconstruction error with respect to the masked information as an abnormality score.

forms of mask: superpixels [36], future frames [37], middle bounding boxes [17], among others.

[17] Anomaly Detection in Video via SelfSupervised and Multi-Task Learning, cvpr21

[36] Superpixel Masking and Inpainting for Self-Supervised Anomaly Detection, bmvc20

SSPCAB, cvpr22 [15, 19, 21, 36, 37, 43, 47, 49, 54, 62, 69, 71]

other applications: denoising (3.3.2.3 Noise2Void, cvpr19 ), de-snowing (SLiDE, eccv22)

Shape of mask vs application types??

3 Code

  1. https://github.com/M-3LAB/awesome-industrial-anomaly-detection

  2. https://github.com/openvinotoolkit/anomalib

    1. Anomalib is a deep learning library that aims to collect state-of-the-art anomaly detection algorithms for benchmarking on both public and private datasets

    2. https://arxiv.org/pdf/2202.08341.pdf

      这个库感觉写的很棒，适合新手快速进步。

  3. https://github.com/EricLee0224/PAD

    1. selected 1-2 representative methods from each of the 8 anomaly detection paradigms

4 Datasets

see PAD, nips23 or Deep industrial image anomaly detection: A survey, machine intelligence research, 24

4.1 MVTec AD

Untitled

  1. MVTec AD -- A Comprehensive Real-World Dataset for Unsupervised Anomaly Detection, JICV21