Gesture recognition based on modified Yolov5s

• DOI: 10.1049/ipr2.12477
• Publication Date: 2022-03-18T12:41:34Z
• Authors: Dunli Hu; Jun Zhu; Jiayu Liu; Jiaju Wang; Xiaoping Zhang
• Keywords: 0202 electrical engineering, electronic engineering, information engineering; 02 engineering and technology

Abstract With the development of artificial intelligence technology, human–computer interaction technology through gestures, images and voices has gradually become a hot topic for discussion. A modified Yolov5s gesture recognition method is proposed in field cooperation by optimizing network structure backbone, CNN replaced Ghostbottleneck module to increase target occlusion rate. Secondly, tensor stitching added output up sampling strengthen reuse image features. Finally, detection ability improved model face complex environment verified on self‐made data set. Experimental results show that, mAP@0.5 (mean average precision) 94.49%, AP (average 94.2%. By comparing algorithm, Yolov4 algorithm,Yolov3 algorithm SSD accuracy been significantly improved, which can fully meet application requirements real‐time gesture‐controlled robots.