Elevating Wafer Defect Inspection with Denoising Diffusion Probabilistic Model

Ping-Hung Wu, Thi Phuong Hoang, Yen-Ting Chou, Andres Philip Mayol, Yu-Wei Lai, Chih-Hsiang Kang, Yu-Cheng Chan, Siou-Zih Lin and Ssu-Han Chen ()
Additional contact information
Ping-Hung Wu: Product Testing Service Office, Nanya Technology Corporation, New Taipei City 243089, Taiwan
Thi Phuong Hoang: Department of Industrial Engineering and Management, Ming Chi University of Technology, New Taipei City 243303, Taiwan
Yen-Ting Chou: Department of Industrial Engineering and Management, Ming Chi University of Technology, New Taipei City 243303, Taiwan
Andres Philip Mayol: Manufacturing Engineering and Management Department, De La Salle University, Manila 0922, Philippines
Yu-Wei Lai: Center for Artificial Intelligence & Data Science, Ming Chi University of Technology, New Taipei City 243303, Taiwan
Chih-Hsiang Kang: Center for Artificial Intelligence & Data Science, Ming Chi University of Technology, New Taipei City 243303, Taiwan
Yu-Cheng Chan: Center for Artificial Intelligence & Data Science, Ming Chi University of Technology, New Taipei City 243303, Taiwan
Siou-Zih Lin: AI Chip Application & Green Manufacturing Department, Industrial Technology Research Institute, Hsinchu 310401, Taiwan
Ssu-Han Chen: Department of Industrial Engineering and Management, Ming Chi University of Technology, New Taipei City 243303, Taiwan

Mathematics, 2024, vol. 12, issue 20, 1-15

Abstract: Integrated circuits (ICs) are critical components in the semiconductor industry, and precise wafer defect inspection is essential for maintaining product quality and yield. This study addresses the challenge of insufficient sample patterns in wafer defect datasets by using the denoising diffusion probabilistic model (DDPM) to produce generated defects that elevate the performance of wafer defect inspection. The quality of the generated defects was evaluated using the Fréchet Inception Distance (FID) score, which was then synthesized with real defect-free backgrounds to create an augmented defect dataset. Experimental results demonstrated that the augmented defect dataset significantly boosted performance, achieving 98.7% accuracy for YOLOv8-cls, 95.8% box mAP for YOLOv8-det, and 95.7% mask mAP for YOLOv8-seg. These results indicate that the generated defects produced by the DDPM can effectively enrich wafer defect datasets and enhance wafer defect inspection performance in real-world applications.

Keywords: wafer defect inspection; generative model; denoising diffusion probabilistic model; You Only Look Once Version 8 (search for similar items in EconPapers)
JEL-codes: C (search for similar items in EconPapers)
Date: 2024
References: View complete reference list from CitEc
Citations:

Downloads: (external link)
https://www.mdpi.com/2227-7390/12/20/3164/pdf (application/pdf)
https://www.mdpi.com/2227-7390/12/20/3164/ (text/html)

Related works:
This item may be available elsewhere in EconPapers: Search for items with the same title.

Export reference: BibTeX RIS (EndNote, ProCite, RefMan) HTML/Text

Persistent link: https://EconPapers.repec.org/RePEc:gam:jmathe:v:12:y:2024:i:20:p:3164-:d:1495416

Access Statistics for this article

Mathematics is currently edited by Ms. Emma He

More articles in Mathematics from MDPI
Bibliographic data for series maintained by MDPI Indexing Manager ().