基于YOLOv5的零件识别轻量化算法

Lightweight Algorithm for Part Recognition Based on YOLOv5

为了解决现有的基于深度学习的零件识别模型参数量过大、检测速度慢、检测精度低的问题,以YOLOv5模型为基础,提出了结合轻量级网络和Transformer的零件识别算法。首先,设计了一种轻量级主干特征提取网络,以减少网络的参数量和计算量,并提升推理速度;其次,将Transformer模块与C3模块融合构成C3TR模块,以增强小目标的检测能力;最后,引入噪音净化模块,通过过滤噪音来提高零件识别模型的准确率。模型的检测平均准确率和平均召回率分别达到了86.7%和85.5%,相较原模型分别提升了和24.2%和17.4%。实验结果表明,改进后的模型在实现模型轻量化的同时,具有更快的检测速度和更高的识别准确率。

In order to solve the problems of excessive parameter amount,slow detection speed and low detection accuracy of existing deep learning-based part recognition models,this paper proposes a part recognition algorithm combining a lightweight network and Transformer based on the YOLOv5 model.Firstly,a lightweight trunk feature extraction network is designed to reduce the number of parameters and computation of the network and to improve the inference speed;secondly,the Transformer module is fused with the C3 module to form the C3TR module to enhance the detection of small targets;finally,the noise purification module is introduced to improve the accuracy of the part recognition model by filtering the noise.The average detection accuracy and average recall of the model reach 86.7%and 85.5%,respectively,which are improved and 24.2%and 17.4%,respectively,compared with the original model.The experimental results show that the improved model has faster detection speed and higher recognition accuracy while achieving model lightweighting.