基于嵌入式终端的YOLOv3算法优化实现

Optimized Rearealization of YOLOv3 Algorithm Based on Embedded Terminal

图像目标识别技术是计算机视觉研究领域的热点问题。然而目前先进的目标检测算法大多基于服务器端训练部署,在如今的移动互联网时代背景下无法做到真正的落地应用。同时考虑到国产化芯片和软件开发环境需求,优化并训练了YOLOv3检测模型,并基于嵌入式终端-百度EdgeBoard边缘AI计算盒进行了模型部署。实验结果充分表明优化后的YOLOv3-MobileNetv1模型对行人、车辆、飞机等多类目标均具有良好的检测识别效果。

Image object recognition technology is a hot issue in the field of computer vision research.However,most of the cur-rent advanced object detection algorithms are based on server-side training and deployment.Under the background of today's mobile Internet era,they cannot be truly applied.At the same time,taking into account the needs of localized chips and software develop-ment environment,the YOLOv3 detection model is optimized and trained,and the model is deployed based on the embedded termi-nal,the Baidu EdgeBoard Edge AI Computing Box.Results of experiment fully show that the optimized YOLOv3-MobileNetv1 mod-el has a good detection and recognition effect on pedestrians,vehicles,airplanes and other types of objects.