摘要详情

ID / 提交时间

161 / 2024-08-31 21:37:54

标题

Wafer Defect Detection Based on YOLO-BA

关键字

Wafer,Defect Detection,Deep Learning,YOLOv10,self-attention mechanism

主题及专题

Special Sessions > SS01 AI for Failure Analysis and Reliability of Integrated Circuit(IC)

状态

终稿

作者

ShuXiaotong / Jiangsu Normal University

XuLin / Jiangsu Normal University

HeZhenzhi / Jiangsu Normal University

ShengLianchao / Jiangsu Normal University

YeGuo / Jiangsu Normal University

LuXiangning / Jiangsu Normal University

摘要

Abstract: With the advancement of the chip industry and manufacturing processes are becoming increasingly sophisticated, the scale and complexity of integrated circuits have significantly increased. This has made wafer manufacturing processes more intricate, leading to a higher probability and variety of wafer defects. To enhance production yield and improve process control, it is crucial to identify defects and address corresponding process issues. This paper presents an improved YOLO-BA wafer detection algorithm based on YOLOv10n. The approach introduces Bottleneck Transformers (BoT) and Adaptive Kernel Convolution (AKConv) into the backbone network, effectively suppressing interference from complex background information while reducing feature redundancy in spatial and channel dimensions. This enables more efficient feature extraction of wafer defects while maintaining model lightweightness. Experimental results show that the improved algorithm achieves a 0.9% higher recall rate, 0.8% higher mAP0.5_0.95, and 0.4% higher mAP0.5 compared to the latest YOLOv10n model, with similar parameter counts and GFLOPs.

Keywords: Wafer, Defect Detection, Deep Learning, YOLOv10, self-attention mechanism