改进YOLOv5s的车辆目标检测研究

Research on Vehicle Target Detection for Improved YOLOv5s

为改善在车辆目标检测中实时性和精确度低的问题,提出一种改进的YOLOv5s算法。引入CBAM(Channel Attention Module)注意力模块,以增强网络对特征的关注度;将CIoU Loss替换为EIoU Loss提高目标边界框的回归速度和定位精度;新增一个小目标检测头,提高模型对小目标的检测能力。实验结果显示:改进后的YOLOv5s算法相较原始算法m AP0.5提高11.4%,mAP0.5:0.95提高了7.3%。在满足实时性的同时,模型检测精度大幅提升。

To address the issues of low real-time performance and accuracy in vehicle object detection,we propose an enhanced YOLOv5s algorithm.We introduce the CBAM(Channel Attention Module)attention mechanism to enhance the network's focus on features.We replace the CIoU Loss with the EIoU Loss to improve the speed of regression and the precision of object localization.Additionally,we add a small object detection head to enhance the model's capability to detect small objects.Experimental results demonstrate that the improved YOLOv5s algorithm achieves a significant improvement over the original algorithm,with a 11.4%increase in mAP at IoU 0.5 and a 7.3%increase in mAP at the range of IoU from 0.5 to 0.95.This enhancement not only maintains real-time performance but also substantially boosts the model's detection accuracy.