轻量YOLOv5在RDK_X3上的实时目标识别

Real-time Object Detection Using Lightweight YOLOv5 on RDK_X3

本文研究探讨轻量化YOLOv5模型在RDK_X3嵌入式平台上的实时目标识别应用。通过对YOLOv5模型进行剪枝和量化,实现模型的轻量化,并在RDK_X3上部署运行。研究比较Ubuntu和Windows系统环境下的部署差异,分析板载应用中的计算量和计算速度数据。采用单目相机进行数据采集,对比单目与双目方案的优劣。实验结果表明,轻量化YOLOv5在RDK_X3上能够实现实时目标识别,在资源受限的嵌入式平台上展现出良好的性能和应用前景。

This study investigates the application of a lightweight YOLOv5 model for real-time object detection on the RDK_X3 embedded platform.The YOLOv5 model is lightened through pruning and quantization,then deployed and run on the RDK_X3.The research compares deployment differences between Ubuntu and Windows system environments,and analyzes computational load and speed data in onboard applications.A monocular camera is used for data collection,comparing the advantages and disadvantages of monocular versus binocular solutions.Experimental results show that the lightweight YOLOv5 can achieve real-time object detection on the RDK_X3,demonstrating good performance and application prospects on resource-constrained embedded platforms.