Improved fabric defect detection algorithm of YOLOv8
Aiming at the problems of time-consuming and labor-intensive traditional fabric defect detection methods and slow speed or low precision of mainstream detection models,YOLOv8 model was improved to improve the performance of fabric defect detection.The Selective Attention module(LSKBlock)is integrated into the YOLOv8 backbone network,which enables the model to automatically learn and decide which information deserves attention and processing,and optimize the model's processing effect on relevant information.The convolutional layer is replaced by deforming convolution(DCN)at the neck to enhance the ability of the network to perceive the deformation of the target,and further improve the feature extraction and localization capability of the network.In addition,a lightweight paradigm(Slimneck)is designed to improve the accuracy and reduce the complexity of the model.Performance evaluation was performed on TILDA and fabric defects datasets.The results showed that the mAP of the im-proved YLOLv8 model on the two datasets reached 88.6%and 92.7%respectively,4.1 and 4.0 percentage points higher than that of the original model,and the detection speed met the requirements in actual production.
NETL
NSTL
万方数据
