基于ShuffleNetv2-YOLOv3模型的静态手势实时识别方法

辛文斌,郝惠敏,卜明龙,兰媛,黄家海,熊晓燕

Static gesture real-time recognition method based on ShuffleNetv2-YOLOv3 model

Wen-bin XIN,Hui-min HAO,Ming-long BU,Yuan LAN,Jia-hai HUANG,Xiao-yan XIONG

表 8 不同主干网络的测试结果

Tab.8 Test results of different backbone networks

网络模型 主干网络 mAP TT/h Ws /MB v /(帧·s−1) YOLOv2 Darknet-19 0.930 3.105 202.4 47 YOLOv3-Tiny Tiny 0.974 1.256 34.8 46 YOLOv3 ResNet-50 0.984 2.821 161.2 40 YOLOv3 Darknet-53 0.990 4.104 246.6 41 YOLOv3 MobileNetv2 0.955 2.051 28.0 37 SSD MobileNetv2 0.882 4.390 24.1 19 YOLOv3 ShuffleNetv2-1.0× 0.992 1.594 15.1 44