Real-IAD: A Real-World Multi-View Dataset for Benchmarking Versatile Industrial Anomaly Detection

Chengjie Wang1,2, Wenbing Zhu3,4, Bin-Bin Gao2, Zhenye Gan2, Jiangning Zhang2, Zhihao Gu1, Shuguang Qian4, Mingang Chen5, Lizhuang Ma1,
1Shanghai Jiao Tong University   2Youtu Lab, Tencent   3Fudan University   4Rongcheer Co., Ltd   5Shanghai Development Center of Computer Software Technology  
Paper Code Hugging Face Hugging Face

Dataset Download

If you are interested in using the dataset, you can download it at Hugging Face Hugging Face. Your access request will be automatically approved. For further questions, please feel free to send emails to realiad4ad@outlook.com.

Dataset Organization

Real-IAD dataset provide three main files, each serving a different purpose to cater to your needs.

  1. realiad_1024: This is a version with a resolution of 1024 × 1024 (downsampled from realiad_raw), with a manageable size of around 53GB. It is tailored to meet the needs of most researchers.
  2. realiad_raw: This is the raw, high-resolution data, with a size of approximately 507GB.
  3. realiad_jsons.zip: This file consists of data splits in JSON format, providing a structured and organized way to access specific portions of the dataset as needed.
Interpolate start reference image.

The correspondence between the code (represented by the folder name) and the defect type:

Code (Folder Name) Defect Type
AK pit
BX deformation
CH abrasion
HS scratch
PS damage
QS missing parts
YW foreign objects
ZW contamination

Collection Pipeline of Real-IAD

The Real-IAD dataset originates from a real production line, encompassing steps such as Material Preparation, Prototype Construction, Data Collection, Annotation, and Cleaning. Material Preparation, Prototype Construction, Data Collection, Annotation are done by staffs from Rongcheer Co., Ltd.

Interpolate start reference image.

Real-IAD Dataset

Interpolate start reference image. Interpolate start reference image. Interpolate start reference image.

BibTeX


      @inproceedings{wang2024real,
        title={Real-iad: A real-world multi-view dataset for benchmarking versatile industrial anomaly detection},
        author={Wang, Chengjie and Zhu, Wenbing and Gao, Bin-Bin and Gan, Zhenye and Zhang, Jiangning and Gu, Zhihao and Qian, Shuguang and Chen, Mingang and Ma, Lizhuang},
        booktitle={Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition},
        pages={22883--22892},
        year={2024}
      }