纺织工程学报2025，Vol.3Issue(4)：70-81,12.
基于自适应YOLOv8-DAC的布匹疵点定位与识别方法
Fabric defect localization and recognition method based on adaptive YOLOv8-DAC
摘要
Abstract
There are problems of complex background textures and diverse defects in fabric defect target detec-tion,and it's also difficult to achieve efficient and accurate recognition and localization of fabric defects with ex-isting deep learning algorithms.This article proposes a fabric defect recognition and localization algorithm YO-LOv8-DAC(Deformable And CARAFE).Firstly,it is to design a feature fusion module C2f DAT(C2f Deform-able Attention Transformer)with adaptive weight adjustment capability.This module can suppress the feature weights of irrelevant information,enhance the weight values of key area feature information,effectively strengthen the feature fusion ability of the model,and enhance the target detection ability.Secondly,in response to the problem of single feature information and loss of feature information caused by the up-sampling process of the model,a lightweight CARAFE up-sampling operator is introduced to replace the original up-sampling lay-er of YOLOv8.This operator can up-sample the feature map by adaptively recombining the kernel,enriching the semantic information of the feature map.Simulation experiments show that compared to YOLOv8,mAP@0.5 of YOLOv8-DAC in identifying fabric defects has been improved by 3.8%,achieving an improvement in the accuracy of fabric defect detection.关键词
疵点检测/特征重组/自适应/目标检测/YOLOv8算法Key words
defect detection/feature recombination/adaptation/target detection/YOLOv8分类
信息技术与安全科学引用本文复制引用
胡经纬,龚闯,江维,聂骏杰,陈振,陈伟..基于自适应YOLOv8-DAC的布匹疵点定位与识别方法[J].纺织工程学报,2025,3(4):70-81,12.基金项目
数字化纺织装备湖北省重点实验室开放课题资助项目(DTL2024013). （DTL2024013）
