Research on Gesture Recognition Based on YOLOv5

The gesture recognition technology based on YOLOv5 adopts the Backbone network to train and learn. The graphical interface developed by python Tkinter module has made the real-time gesture recognition of videos with high precision and frame rate achievable. In terms of the testing dataset, the overall average recognition precision of model can be up to 99.6%, which is up to the precision of current mainstream deep learning target detection algorithms. Moreover, it is superior to those algorithms in terms of the recognition speed. From the experimental results, it can be concluded that YOLOv5 takes account of both precision and speed requirements of recognition, which has an advantage in dealing with the natural interaction information. Thus the technology provided by this paper proposes effective means for the human-computer interaction techniques.