A Lightweight Object Detection Network for Industrial Robot Based YOLOv5

Abstract

With the rapid development of deep learning, it has been widely applied in fields such as computer vision, natural language processing, and robotics. Despite the superior performance of deep learning in object detection, most industrial vision robots still rely on traditional object detection methods due to computational constraints of robot controllers. In order to improve the performance of object detection for vision robots, a lightweight 3D object detection network based on You Only Look Once version 5 (YOLOv5) is proposed for satisfying industrial production. YOLOv5 work to efficiently object detection using deep convolutional networks. In terms of model deployment, we adopt a novel OpenVINO-based model deployment approach. The OpenVINO framework significantly enhances the inference speed of models by model optimization and compression. Our model achieves a 70% reduction in inference time compared to the baseline model on CPU. The effectiveness of the proposed method is demonstrated through experiments.