重庆科技大学学报（自然科学版）2025，Vol.27Issue(3)：86-95,10.DOI:10.19406/j.issn.2097-4531.2025.03.009
基于改进YOLOv7的手机屏幕缺陷检测算法
Mobile Phone Screen Defect Detection Algorithm Based on Improved YOLOv7
摘要
Abstract
Aiming at the problem that traditional mobile phone screen defect detection methods are not accurate enough and are prone to miss small target defects,a mobile phone screen defect detection algorithm based on im-proved YOLOv7,named BPC-YOLOv7,is proposed.Firstly,the BiFormer attention mechanism is used to adap-tively assign the most relevant key-value pairs to each query,thereby realizing content-aware sparse patterns and ef-fectively improving the accuracy of feature extraction.Secondly,the PConv+PWConv module is introduced to re-duce the number of parameters and the amount of calculation of the model.Finally,the CARAFE upsampling oper-ator is employed to dynamically generate adaptive kernels to significantly improve the receptive field of the model,thereby better capturing detailed features.In comparison with the YOLOv7 algorithm the experimental results indi-cate that,the average precision of the BPC-YOLOv7 algorithm is increased from 88.4％to 96.1％,and the number of parameters is reduced from 37.6×106 to 33.1×106,which meets the needs of industrial scenarios for mobile phone screen defect detection.关键词
YOLOv7算法/手机屏幕/缺陷检测/BiFormer注意力机制/CARAFE上采样算子Key words
YOLOv7 algorithm/mobile phone screen/defect detection/BiFormer attention mechanism/CARAFE upsampling operator分类
信息技术与安全科学引用本文复制引用
饶宇锋,唐海,张彬,徐洪胜,冯立..基于改进YOLOv7的手机屏幕缺陷检测算法[J].重庆科技大学学报（自然科学版）,2025,27(3):86-95,10.基金项目
湖北省科技厅国际合作交流项目"国产区块链体系架构与共识算法及其在汽车绿色低碳化评价监测科研的应用"(2023EHA018) （2023EHA018）
