G-YOLOX: A Lightweight Network for Detecting Vehicle Types

In recent years, vehicle type detection has had an important role in traffic management. A lightweight network based on multiscale ghost convolution called G-YOLOX is designed this paper. It suitable for practical applications embedded device. Specifically, 3 × convolutions and id="M2"> 5 id="M3"> 7 are combined to fully utilize different feature information. series of linear transformations was generate maps ensure that the lightweight. Moreover, a dataset images showing vehicles city environment established. Altogether, 20,000 road scene were collected, seven categories identified. Extensive experiments with benchmark datasets VOC2007 VOC2012 our demonstrate superiority proposed over original YOLOX. The can achieve nearly invariable mean average precision 0.5, while size weight file decreased by 40% number parameters 67% compared YOLOX network.