Human body key point detection method based on context information and joint embedding

Abstract

The invention discloses a human body key point detection method based on context information and joint embedding. The method comprises the following steps: S1, obtaining training data; S2, building a human body key point detection model; S3, constructing a human body key point real tag heat map and a joint embedded value tag heat map; S4, training the human body key point detection model by using the training data to obtain a trained human body key point detection model; S5, evaluating the human body key point detection model by using verification data, and selecting an optimal model; S6, repeating the steps S4 to S5 until all the training data are trained, and obtaining an optimal model; and S7, performing human body key point prediction by using the optimal model. According to the method, modeling is carried out on internal correlation of pixels in a feature map through a context fusion module, the prediction progress of key point pixel positions is improved, and meanwhile correct matching between different human body key points is guaranteed by constructing joint embedding loss, reducing training difficulty and increasing convergence speed.

Description

technical field [0001] The invention relates to the technical fields of computer vision and deep learning, in particular to a method for detecting key points of a human body based on context information and joint embedding. Background technique [0002] Human key point detection is an important branch in the field of computer vision, which plays a key role in the research of human behavior analysis, gesture recognition and tracking and other tasks. Human body key point detection is mainly to detect some important key points of the human body (such as human body joints or other important parts) in the image or video, that is, given an input image containing at least one pedestrian, the output image contains all The position coordinates of all the key points of pedestrians can be obtained by pairing the coordinates of the output key points to obtain the pose information of all people in the image. [0003] Traditional methods mainly include graph models and graph structures, ...

AI Technical Summary

Problems solved by technology

Therefore, the top-down detection method can be regarded as a two-stage serial mode of pedestrian detector plus single-person key point detection. This is not an end-to-end model, and the detection accuracy of the latter stage largely depends on the previous stage. The output results of the first stage, in some dense and complex scenes, the pedestrian detector often has serious missed detection

In addition, its detection time is affected by the number of pedestrians in the image, the detection speed is difficult to guarantee, and it is difficult to popularize and apply in actual scenes

[0005] The bottom-up detection method does not need to use a pedestrian detector, directly predicts the key points of all pedestrians in the image, and then groups and pairs all the key points. Due to its relatively fast detection speed, the bottom-up detection method is There is a lot of room for development in actual application scenarios, but the detection accuracy of the current bottom-up method is still subject to many limitations in practical applications, the detection speed is not fast enough, and the detection accuracy is not high

Images