AI and Advanced 3D Imaging for Real-Time Anomaly Detection
Project Description
This project focuses on developing systems and algorithms for real-time anomaly detection in applications. The goal is to improve quality control and reduce defects in production processes. Traditional inspection methods rely on 2D imaging, which may fail to detect subtle geometric irregularities, or manual inspection, which is labor-intensive. Meanwhile, 3D imaging is too slow to be useful in practice. To address these challenges, this project develops an integrated solution that incorporates 3D signals into the anomaly detection process by tailoring advanced 3D optical imaging technologies and AI models to analyze the data with high speed and high accuracy.
Our approach investigates recent advances in 3D computer vision, such as GaussinSplats and Feed-Forward 3D models, and will study practical anomaly detection scenarios in the online and few-shot settings. As with all AI models, the key to success is the data quality. To acquire high-quality data, Dr. Zhang’s lab has developed advanced 3D optical imaging technologies with both hardware and software expertise.
We will aim for a system that operates in real-time to satisfy practical manufacturing constraints, e.g., in industries like automotive, pharmacy, and electronics. Ultimately, the system will minimize waste and lower inspection costs through intelligent and automated visual inspection systems.
Start Date
June 1, 2026
Postdoc Qualifications
|
- Ph.D. in Computer Vision, Machine Learning, or a related field with strong experience in 3D data processing (point clouds, meshes). - A solid publication record in top-tier conferences (e.g., CVPR, ICCV, ECCV, NeurIPS) related to computer vision, deep learning, or 3D processing. - The ideal candidate should be an independent, self-motivated researcher with strong problem-solving and communication skills - An engineering background would be a plus. |
Co-advisors
- Raymond Yeh (rayyeh@purdue.edu), Department of Computer Science, https://raymond-yeh.com/
- Song Zhang (zhan2053@purdue.edu), School of Mechanical Engineering, www.xyztlab.com
Bibliography
- Yang, Chiao-An, Kuan-Chuan Peng, and Raymond A. Yeh. "Toward Long-Tailed Online Anomaly Detection through Class-Agnostic Concepts." Proc. ICCV, 2025
- Zhou, Zheyuan, et al. "R3d-AD: Reconstruction via diffusion for 3d anomaly detection." Proc. ECCV, 2024
- Zhang, Haomeng, Chiao-An Yang, and Raymond A. Yeh. "Multi-object 3D grounding with dynamic modules and language-informed spatial attention." Proc. NeurIPS, 2024
- J. Girard and S. Zhang, “Fast error detection method for additive manufacturing process monitoring using structured light three dimensional imaging technique,“ Optics and Lasers in Engineering, 184, 108609 (2025)
- L. Chen and S. Zhang, “Electrically tunable lens assisted absolute phase unwrapping for large depth-of-field 3D microscopic structured-light imaging,” Optics and Lasers in Engineering, 174, 107967 (2024)