Training method of SMPL parameter prediction model, server and storage medium

Abstract

The invention discloses a training method of an SMPL parameter prediction model, a server and a storage medium. The method comprises the steps of obtaining a sample picture; inputting the sample picture into an attitude parameter prediction model to obtain an attitude prediction parameter; inputting the sample picture into a morphological parameter prediction model to obtain a morphological prediction parameter; according to the posture prediction parameters and the form prediction parameters, constructing a human body three-dimensional model through an SMPL model; calculating model predictionloss according to the SMPL prediction parameters and / or the human body three-dimensional model in combination with the annotation information of the sample picture; and reversely training an attitudeparameter prediction model and a morphological parameter prediction model according to the model prediction loss. In the embodiment of the invention, the sample picture is directly used as model input for model training, and a model for extracting the human body information in the picture does not need to be independently trained, so that the complexity of model training is reduced, and the efficiency of model training is improved.

Description

technical field [0001] The embodiment of the present application relates to the field of computer vision, and in particular to a training method, server and storage medium of an SMPL parameter prediction model. Background technique [0002] 3D human body reconstruction is one of the important topics in computer vision research, and has important application value in the fields of virtual reality (VR, Virtual Reality), human animation, games and so on. [0003] In the related art, a Skinned Multi-Person Linear (SMPL, Skinned Multi-Person Linear) model is used to perform three-dimensional human body reconstruction on a human body in a two-dimensional image. In a 3D human body reconstruction method, the human body information extraction model is first used to extract human body information such as 2D joint points, 3D joint points, 2D human body segmentation maps, and 3D voxels in the 2D image, and then the extracted human body The information is input into the parameter predic...

AI Technical Summary

Problems solved by technology

[0005] The embodiment of the present application provides a training method, server and storage medium of an SMPL parameter prediction model, which can solve the problem that the model training process in the related art is complicated and takes a lot of time

Images