Model training method, human body posture detection method and device, equipment and medium

Abstract

The invention discloses a model training method, a human body posture detection method and device, equipment and a medium, and the method comprises the steps: inputting a corresponding image trainingsample under the current iteration into a current posture detection network model, obtaining a corresponding first loss function under the current iteration according to the data form of the image training sample, carrying out reprojection processing on a current output result of the attitude detection network model according to the obtained camera parameters, acquiring a corresponding second lossfunction under current iteration based on a reprojection processing result, and performing back propagation on the attitude detection network model based on a fitting loss function formed by the first loss function and the second loss function to obtain an attitude detection network model for the next iteration. By utilizing the method, the collection difficulty of training samples is reduced, the easiness in implementation of network training is ensured, and meanwhile, the whole model training is completed in an image domain, so that the stability and the rapid convergence of the model training are facilitated.

Description

technical field [0001] The embodiments of the present invention relate to the technical field of spatial position detection, and in particular to a model training method, a human body posture detection method, a device, a device, and a medium. Background technique [0002] With the rapid development of artificial intelligence technology, artificial neural networks have been widely used. Artificial neural network, also known as neural network, is a model that simulates the structure of brain synaptic connections for information processing. In the field of three-dimensional space position detection, neural network technology can be used to construct a pose estimation neural network model for human three-dimensional pose detection. [0003] The detection of the three-dimensional posture of the human body has become an important problem in the field of computer vision. This type of technology can be considered as the basis for computers to understand human behavior. It has bro...

AI Technical Summary

Problems solved by technology

[0004] However, the existing pose detection neural network model for human body 3D pose detection has the following problems: 1) Due to the lack of effective natural scene 3D human body labeling methods in the industry, there is currently a lack of high-quality unconstrained scene labeling data; 2) The model has problems with convergence difficulties

As a result, it is impossible to accurately detect the three-dimensional posture of the human body in natural scenes based on the existing posture detection neural network model.

Images