in 3D human body pose and shape estimation for applications such as the metaverse,gaming,and virtual reality. First,image features are extracted and input into a motion?
continuity attention module to better calibrate thetime sequence range that requires attention. Then,a real - time feature attention integration module is used to effectively?
combine the feature representations of the current frame and past frames. Finally,the human parameter regression network is used to obtain the finalresults,and a graph convolutional generative adversarial network is used to determine whether the model comes from real human motiondata. Compared with previous methods based on real -
time video streams, the proposed method reduces the acceleration error by anaverage of 30% on mainstream datasets,while reducing the network parameters and comput－ational complexity by 65% . The proposedmethod achieves a 3D human body pose and shape estimation speed of 55 ~ 60 frames per second in practical tests,providing?
better userexperience and higher application value for applications such as the metaverse,gaming,and virtual reality.