基于改进YOLOv5s汽车驾舱遗忘物检测

Based on the improved YOLOv5s car cockpit forgetting detection

摘要:

针对目前汽车驾舱内遗忘物检测精度不高的问题,提出一种基于改进的YOLOv5s汽车驾舱遗忘物的检测方法。该检测方法将YOLOv5s作为基础网络,在此基础上进行改进。首先在主干网络的尾部引入SE注意力模块,加强模型对通道信息的关注提升目标检测性能；其次改进空间金字塔池化模块,将原有的SPPF模块改进为SPPCSPC模块,通过增加一点计算量来进一步提升检测模型的精度；最后同时引入GSConv层,S能够缓解DSC(深度可分离卷积)的缺陷,并充分利用DSC的优势,在小目标检测方面取得明显的提升效果,既保证了语义信息又平衡了模型的准确性,也提升了检测速度。通过训练结果说明,改进后的网络与原YOLOv5s网络相比,其平均精度均值mAP提高了2%,查准率提升了3.5%。改进后的网络具有良好的提升效果,表明了该方法的有效性。

Abstract:

Aiming at the problem that the detection accuracy of amnesia in the cockpit of automobiles is not high, a detection method based on the improved YOLOv5s automotive cockpit forgetting is proposed. The detection method uses YOLOv5s as the basic network and improves on this basis. Firstly, the SE attention module is introduced at the tail of the backbone network to strengthen the model"s attention to channel information and improve the object detection performance. Secondly, the spatial pyramid pooling module is improved, and the original SPPF module is improved to the SPPCSPC module, which further improves the accuracy of the detection model by adding a little calculation. Finally, the GSConv layer is introduced at the same time, which balances the accuracy of the model and the detection speed while ensuring the semantic information. Experimental results show that compared with the original YOLOv5s network, the average accuracy mean mAP of the improved network is increased by 2%, and the accuracy is increased by 3.5%. The improved network has a good improvement effect, which shows the effectiveness of the method.