Research on Wafer Reference Mark Recognition Method Based on Lightweight Improved YOLOv5
Abstract
hardly meet the requirements of precision positioning. This paper proposes a lightweight method based on YOLOv5s and structured pruning, which achieves the balance between model compression and accuracy through simulated degraded data, transfer learning and structured pruning. Experimental results show that the proposed method effectively reduces the model parameters and computational cost while
maintaining high-precision recognition performance, has significant advantages in degraded scenarios, and can be efficiently deployed on
ordinary CPU devices.
Keywords
Full Text:
PDFReferences
[1] Lu P X, Yang K M, Lu S, et al. Alignment mark location algorithm for wafer bonding[J]. Chinese Journal of Scientific Instrument, 2021,
42(11): 220-229.
[2] Hu Z Q, Wu Y Q. Review of semiconductor wafer defect detection methods based on machine vision[J]. Journal of Image and Graphics,
2025, 30(1): 25-50.
[3] Cao Y Y. Research on wafer surface defect detection system based on deep learning[D]. North China University of Water Resources and
Electric Power, 2023.
[4] Hu W, Zhao J M, Li D A. Lightweight laser chip defect detection algorithm based on improved YOLOv7-Tiny[J]. Journal of Taiyuan
University of Technology, 2025, 56(1): 137-147.
[5] Tang M H, He M Y, Liu X Q, et al. An improved wafer surface defect detection method based on YOLOv8: Chinese Patent[P]. 2023-11-15.
DOI: http://dx.doi.org/10.70711/frim.v4i5.9373
Refbacks
- There are currently no refbacks.
