Motion recognition method and computing device

Abstract

The invention discloses an action recognition method and computing equipment. The action recognition method comprises the steps that image data are sent to an action recognition model to obtain an action recognition result, the action recognition model is a new OfficientNet recognition network model constructed based on a basic network structure of the OfficientNet, and the new OfficientNet recognition network model modifies at least one of a module, a layer and an operation direction in the basic network structure of the OfficientNet. According to the training method, one of a plurality of new OfficientNet recognition network models is selected as an action recognition model, so that in the action recognition process, the operation speed can be increased, and the operation precision can be ensured.

Description

technical field [0001] The present invention relates to the technical field of image processing, and more specifically, to an action recognition method and a computing device. Background technique [0002] In recent years, due to the development of educational video broadcasting, an educational intelligent recording system is needed. The intelligent recording system uses high-definition cameras to collect images of teachers and students. Through video image analysis, it can switch between various teaching scenes and automatically follow up with teachers and students. , and recognize the action of the subject. [0003] The intelligent recording system needs to detect and track the teacher in the video, and recognize the teacher's lecture actions. However, conventional techniques based on gesture recognition methods are time-consuming, such as OpenPose, DensePose, etc. In actual teacher lecture recognition methods, the processing speed is often too slow to meet the needs of p...

AI Technical Summary

Problems solved by technology

However, conventional techniques based on gesture recognition methods are time-consuming, such as OpenPose, DensePose, etc. In actual teacher lecture recognition methods, the processing speed is often too slow to meet the needs of practical applications

[0004] For example, the OpenPose algorithm is used to detect the teacher in the image and recognize the lecture action. The OpenPose algorithm is used to extract the key points of the human body in the input image. The algorithm adopts a bottom-up strategy, first extracting the position of the key points in the image, and then The learned key point relationship calculates the human skeleton information, and the position of the teacher in the image can be calculated according to these key point information, which solves the problem of teacher detection, but in the actual teacher lecture recognition method, the processing speed of the algorithm is often slower Slow, unable to meet the needs of practical applications

Images