Joint point detection method and device, storage medium and electronic device

Abstract

The invention discloses a joint point detection method and device, a storage medium and an electronic device. The joint point detection method comprises the steps of obtaining a human body image in ato-be-processed image; generating a detection joint point heat map of the human body image by using a generator of the generative adversarial network model; and determining joint positions in the to-be-processed image according to the detection joint heat map of the human body image. According to the technical scheme provided by the embodiment of the invention, high-precision joint point detectionis realized.

Description

technical field [0001] Embodiments of the present invention relate to the field of joint point detection, and in particular, to a joint point detection method, device, storage medium, and electronic equipment. Background technique [0002] Human body pose estimation technology has been applied in the industry, such as the detection and recognition of human body joints in the Kinect game console. Generally speaking, this technology is relatively new and has a large potential application market, such as robotics, augmented reality applications, virtual reality applications, physical education industries, etc. [0003] The existing human joint point detection method is to use the convolutional neural network to establish a joint point detection model, and use the joint point detection model to generate a joint point heat map of each joint point. The value of each position in the heat map represents that the joint point appears in the The probability of the position, the positi...

AI Technical Summary

Problems solved by technology

However, the existing human joint point detection method uses the gradient descent method when training the joint point detection model, which requires the cost function to be differentiable, and the function to obtain the position coordinates of the maximum value in the joint point heat map is not differentiable. Therefore, the present The cost function adopted by some human joint point detection methods is the difference between the joint point heat map generated by the joint point detection model and the value of the corresponding coordinate position of the real joint point heat map

The cost function cannot reflect the distance relationship between the joint point position in the joint point heat map generated by the joint point detection model and the real joint point position, making the joint point detection accuracy not high

Images