If you are interested in using the dataset, please send an email to realiad4ad@outlook.com from your affiliation email address. In your email, kindly provide us with the necessary information for the application. Once we receive your email, we will respond with a download link and password. Thank you for your interest and cooperation. We hope you have a pleasant experience using our dataset! A recommended application email format is provided in here.
The Real-IAD D³ dataset comprises 20 product categories and 10 defect types. Each sample includes high-resolution RGB images, micrometer-level 3D point clouds, and pseudo-3D images generated through photometric stereo, all captured in synchronization. The dataset features defect area proportions ranging from 0.46% to 6.39% and provides pixel-level annotations.
The correspondence between the code (represented by the folder name) and the defect type:
| Code (Folder Name) | Defect Type |
|---|---|
| HS | scratch |
| KD | hole |
| QK | porosity |
| KS | impact damage |
| AK | pit |
| BMLL | exposed surface |
| PS | damage |
| JSYW | metallic contaminant |
| YW | oil stain |
| BX | deformation |
The Real-IAD D³ dataset originates from a real production line, encompassing steps such as Material Preparation, Defect Manufacturing, Image Acquisition Setup, Labeling and Cleaning, Annotation are done by staffs from Rongcheer Co., Ltd.
Click here to obtain the hardware parameter report for data collection.
Data Collection Demonstration Video
@article{zhu2025real,
title={Real-IAD D3: A Real-World 2D/Pseudo-3D/3D Dataset for Industrial Anomaly Detection},
author={Zhu, Wenbing and Wang, Lidong and Zhou, Ziqing and Wang, Chengjie and Pan, Yurui and Zhang, Ruoyi and Chen, Zhuhao and Cheng, Linjie and Gao, Bin-Bin and Zhang, Jiangning and others},
journal={arXiv preprint arXiv:2504.14221},
year={2025}
}