Image action recognition method, device and electronic equipment

Abstract

Embodiments of the present invention relate to an image action recognition method, device, and electronic equipment. The method includes: acquiring an action image to be identified; performing feature extraction processing on the action image to obtain an image feature vector; inputting the image feature vector into an N×M The convolutional layer performs fusion processing to obtain image fusion features; the image fusion features are input to the classification layer, and based on the image fusion features, action classification and recognition are performed on action images. In this process, while trying not to affect the model reasoning speed, it can greatly improve the model's recognition accuracy of image actions.

Description

technical field [0001] Embodiments of the present invention relate to the field of computer technology, and in particular, to an image action recognition method, device, and electronic equipment. Background technique [0002] Action recognition is to classify videos containing human actions, and it plays an important role in applications such as video surveillance and human-computer interaction, so it has been extensively studied. [0003] In recent years, compared with traditional RGB video recognition methods, skeleton-based action recognition has attracted increasing attention due to its strong adaptability to dynamic environments and complex backgrounds. Early skeletal action recognition methods based on deep learning manually constructed skeletal data into joint coordinate vector sequences or pseudo-images, and input them into Recurrent Neural Network (RNN) or Convolutional Neural Network (CNN for short) generate forecasts. However, representing skeleton data as a seq...

AI Technical Summary

Problems solved by technology

[0006] The present application provides an image action recognition method, device and electronic equipment to solve the problem that the action recognition of bones is not very accurate in the adaptive graph convolutional network model in the prior art

Images