Key point prediction method and device, electronic device and storage medium

Abstract

The embodiment of the invention discloses a key point prediction method and device, an electronic device and a storage medium. The method comprises: a target area including a to-be-identified person in a video frame is obtained; the target area is inputted into any neural network unit in a neural network and a key point hotspot map is generated, wherein the neural network includes a plurality of cascaded neural network units, each neural network unit includes a long-term and short-term memory sub network for screening a memory feature according to the target area and an output result from thelong-term and short-term memory sub network corresponding to the previous-stage neural network unit of the current neural network unit, and the memory feature is used for determining the key point hotspot map; and on the basis of the key point hotspot, location information of a body key point in the video frame is determined. Therefore, quick and high-precision key point prediction of a video is realized; and the accuracy of the key point prediction is improved.

Description

technical field [0001] Embodiments of the present invention relate to image recognition technology, and in particular to a method, device, electronic device, and storage medium for predicting key points of a human body. Background technique [0002] The prediction of key points of the human body is an important research topic in the field of visual analysis of human motion, and is widely used in fields such as augmented reality, computer animation, and automatic photo processing. Most of the traditional predictions of key points of the human body are based on manually designed graphical models, which are limited by the performance of the models, resulting in low prediction accuracy. [0003] With the development of convolutional neural networks, the prediction technology of key points of the human body has also been rapidly developed, but the current predictions are all static image predictions, which cannot integrate temporal features, and the prediction errors of static ke...

AI Technical Summary

Problems solved by technology

Most of the traditional predictions of key points of the human body are based on manually designed graphical models, which are limited by the performance capabilities of the models, resulting in low prediction accuracy

[0003] With the development of convolutional neural networks, the prediction technology of key points of the human body has also been developed rapidly, but the current predictions are all static image predictions, which cannot integrate time-series features, and the prediction errors of static key point predictions lead to video. The key points of the human body are incoherent in each consecutive video frame, and the prediction accuracy is reduced

Images