Human body joint point detection model training method, human body joint point detection method and storage medium terminal

Abstract

The invention discloses a human joint point detection model training method and detection method, and a storage medium terminal, and the model training method comprises the steps: obtaining a to-be-trained data set, and training a preset structure training model based on the to-be-trained data set, so as to obtain a human joint point detection model; the preset structure training model comprises a convolution module, a plurality of Inception module groups and an output module which are connected in sequence; the number of the Inception module groups is not more than four; each Inception module group comprises a plurality of Inception Modules, and each Inception Module comprises four branches which are connected in parallel; wherein the first branch comprises a 1 * 1 convolution layer, the second branch comprises a 1 * 1 convolution layer and a 5 * 5 convolution layer which are connected in sequence, the third branch comprises a 1 * 1 convolution layer and a 3 * 3 convolution layer which are connected in sequence, and the fourth branch comprises a 3 * 3 average pooling layer and a 1 * 1 convolution layer which are connected in sequence. According to the model training method, the detection process has high precision and high real-time speed.

Description

technical field [0001] The invention relates to the technical field of image recognition, in particular to a training method and a detection method for a human joint point detection model, a storage medium and a terminal. Background technique [0002] Human gesture recognition has broad application prospects and important use values ​​in many fields, such as human-computer interaction systems, film production, and medical aided diagnosis. Especially in the typical application scenarios of smart care for aging-oriented applications, when the elderly fall, it is of great significance to detect in real time, issue an alarm, and deal with them in time. Human joint detection is an essential step in the process of human gesture recognition, but the existing human joint detection methods generally have the problems of large amount of calculation, low detection accuracy, and slow real-time speed. SUMMARY OF THE INVENTION [0003] The technical problem to be solved by the present ...

AI Technical Summary

Problems solved by technology

[0003] The technical problem to be solved by the present invention is that the existing human body joint point detection methods generally have a large amount of calculation, low detection accuracy and slow real-time speed.

Images