M2AD Dataset

A Multi-View and Multi-Illumination Dataset For Anomaly Detection


Read More
M2AD Dataset | PAPER M2AD Dataset

PAPER

Visual Anomaly Detection under Complex View-Illumination Interplay: A Large-Scale Benchmark

Yunkang Cao*, Yuqi Cheng*, Xiaohao Xu, Yiheng Zhang, Yihan Sun, Yuxiang Tan, Yuxin Zhang, Xiaonan Huang, Weiming Shen
Huazhong University of Science and Technology * Equal contribution

The practical deployment of Visual Anomaly Detection (VAD) systems is hindered by their sensitivity to real-world imaging variations, particularly the complex interplay between viewpoint and illumination which drastically alters defect visibility. Current benchmarks largely overlook this critical challenge. We introduce Multi-View Multi-Illumination Anomaly Detection (M2AD), a new large-scale benchmark comprising 119,880 high-resolution images designed explicitly to probe VAD robustness under such interacting conditions. By systematically capturing 999 specimens across 10 categories using 12 synchronized views and 10 illumination settings (120 configurations total), M2AD enables rigorous evaluation. We establish two evaluation protocols: M2AD-Synergy tests the ability to fuse information across diverse configurations, and M2AD-Invariant measures single-image robustness amidst realistic view-illumination effects. Our extensive benchmarking shows that state-of-the-art VAD methods struggle significantly on M2AD, demonstrating the profound challenge posed by view-illumination interplay. This benchmark serves as an essential tool for developing and validating VAD methods capable of overcoming real-world complexities.

Cite Us

If you use this dataset in your research, please cite the following paper:

@inproceedings{M2AD,
    title={Visual Anomaly Detection under Complex View-Illumination Interplay: A Large-Scale Benchmark},
    author={Cao, Yunkang and Cheng, Yuqi and Xu, Xiaohao and Zhang, Yiheng and Sun, Yihan and Tan, Yuxiang and Zhang, Yuxin and Huang, Xiaonan and Shen, Weiming},
    booktitle={https://arxiv.org/abs/2505.10996},
    year={2025},
}

THE DATASET

Data Acquisition

Alt text

Ten Object Classes

BirdCarCubeDiceDoll
Alt textAlt textAlt textAlt textAlt text
HolderMotorRingTeapotTube
Alt textAlt textAlt textAlt textAlt text

Multi-View

BirdCarCubeDiceDoll
Alt textAlt textAlt textAlt textAlt text
HolderMotorRingTeapotTube
Alt textAlt textAlt textAlt textAlt text

Multi-Illumination

BirdCarCubeDiceDoll
Alt textAlt textAlt textAlt textAlt text
HolderMotorRingTeapotTube
Alt textAlt textAlt textAlt textAlt text

Multi-View & Multi-Illumination

BirdCarCubeDiceDoll
Alt textAlt textAlt textAlt textAlt text
HolderMotorRingTeapotTube
Alt textAlt textAlt textAlt textAlt text

LEADERBOARD

MethodBirdCarCubeDiceDollHolderMotorRingTeapotTubeMean
CDO70.6/74.1/90.176.8/65.2/77.972.2/64.9/72.493.0/82.0/82.269.9/64.0/74.496.0/78.1/72.983.7/69.7/94.091.6/84.9/88.892.6/79.8/92.696.5/81.8/93.784.3/74.4/83.9
RD++90.3/70.2/79.885.0/68.2/75.683.1/74.6/80.798.4/89.4/85.666.8/65.9/85.499.1/87.8/81.092.2/87.9/94.995.5/90.9/77.291.3/86.0/91.792.1/81.2/90.989.4/80.2/84.3
MSFlow85.0/62.0/71.467.9/55.9/67.466.0/57.8/58.776.8/69.4/77.056.4/55.1/68.998.0/76.6/59.686.0/61.4/86.774.7/72.4/83.983.0/63.9/77.389.0/67.3/84.178.3/64.2/73.5
Dinomaly75.1/74.9/86.986.7/75.1/78.382.3/77.8/86.098.1/93.0/85.774.4/72.6/89.099.7/85.8/90.095.4/85.4/94.291.2/87.3/77.899.9/94.6/94.397.2/83.3/77.090.0/83.0/85.9
INP-Former80.0/67.2/84.158.1/53.9/72.177.9/74.5/80.693.3/83.7/87.772.5/73.7/85.899.2/76.4/81.083.7/61.1/91.975.5/71.7/91.491.6/79.1/92.478.0/64.1/85.971.0/70.5/85.3

Submit Your Results

Send us the performance of your method on our dataset, and we will add your results to our leaderboard!

Please send an e-mail to yuqicheng@hust.edu.cn with subject “results of {Your method name} on M2AD” and the following info:

  • The full name/s of your method/s.
  • A link to a published paper describing it/them.
  • A csv or xlsx file with detailed results.

Feel free to ask us any questions!

Yunkang Cao

Yunkang Cao

caoyunkang@ieee.org

Yuqi Cheng

Yuqi Cheng

yuqicheng@hust.edu.cn (Contact me first!)

Xiaohao Xu

Xiaohao Xu

xiaohaox@umich.edu

Yiheng Zhang

Yiheng Zhang

yihengzhang@hust.edu.cn

Yihan Sun

Yihan Sun

yihansun@hust.edu.cn

Yuxiang Tan

Yuxiang Tan

yuxiangtan@hust.edu.cn

Yuxin Zhang

Yuxin Zhang

zyx_hust@hust.edu.cn


You Can find us here