A gesture recognition model training method, gesture recognition method and system

Abstract

The invention discloses a gesture recognition model training method, gesture recognition method and system, wherein the training method includes collecting gesture picture samples in various scenarios, randomly cropping the gesture samples to obtain new gesture samples, and the new gesture samples as the sample set; construct the Light YOLO network, use the sample set to train the Light YOLO network, and obtain the first Light YOLO network; add a selective-dropout after each convolutional layer of the first Light YOLO network layer to obtain the second Light YOLO network, use the sample set to train and converge the second Light YOLO network, and then cut the channel to obtain the gesture recognition model. The invention improves the detection performance of the network for gestures with smaller resolution. The gesture recognition method of the present invention has high accuracy and good real-time performance. At the same time, the system of the present invention can directly obtain the recognition result from the picture, and can perform end-to-end optimization.

Description

technical field [0001] The invention belongs to the technical field of computer vision, and more specifically relates to a gesture recognition model training method, a gesture recognition method and a system. Background technique [0002] As one of the most natural body languages, gestures can be applied in the field of human-computer interaction to make the interaction process more natural. Among them, the recognition of human hands is the focus of current research in the field of human-computer interaction. Scholars at home and abroad have carried out a lot of research on vision-based gesture recognition technology. [0003] Traditional gesture recognition systems generally perform gesture segmentation first to obtain gesture regions, then extract gesture features, and finally use gesture features for classification. Traditional methods require manual design of features, such as color features, HOG features, etc. These features have poor generalization ability, and differ...

AI Technical Summary

Problems solved by technology

[0005] In view of the above defects or improvement needs of the prior art, the present invention provides a gesture recognition model training method, gesture recognition method and system, thereby solving the problems of low accuracy and poor real-time performance in the existing gesture recognition methods in complex scenes question

Images