Human body posture estimation method based on articulation point difficult case mining

Abstract

The invention provides a human body posture estimation method based on articulation point difficult case mining, and belongs to the field of human body posture estimation. The method comprises the following steps: acquiring a public data set of human body posture estimation; according to the method, a CMU-Pose network is improved, a backbone network in the CMU-Pose network is replaced by a ResNet network in which an attention mechanism is introduced, so that useful features are improved, features with small use are inhibited, and an articulation point difficult case mining algorithm is introduced, so that the extraction effect of articulation points, such as wrists, ankles and the like, which are relatively flexible and relatively difficult to estimate is improved; training the improved CMU-Pose network by using the images in the public data set; and inputting a to-be-detected image into the trained and improved CMU-Pose network to obtain a human body posture estimation result. By adopting the method and the device, the prediction accuracy of joint points which are relatively difficult to predict and relatively flexible can be improved.

Description

technical field [0001] The invention relates to the field of human body posture estimation, in particular to a human body posture estimation method based on difficult example mining of joint points. Background technique [0002] Two-dimensional human pose estimation methods are roughly divided into two categories: top-down multi-person two-dimensional human pose estimation, and bottom-up multi-person two-dimensional human pose estimation. The top-down approach refers to first using an object detection algorithm to detect each person in the image. Then a single-person 2D human pose estimation is performed for each detection frame. Finally, the multi-person two-dimensional human pose estimation results are obtained. The top-down method is more accurate, but its effect is very dependent on the performance of the detector, and its running time is proportional to the number of people in the picture. The bottom-up human pose method has lower accuracy than the top-down method, b...

AI Technical Summary

Problems solved by technology

The top-down method has higher accuracy, but its effect is very dependent on the performance of the detector, and its running time is proportional to the number of people in the picture

Although the accuracy of the bottom-up human pose method is lower than that of the top-down method, its network running time remains basically the same as the number of people in the picture increases.

However, this method has the problems of network structure and poor extraction accuracy for more flexible joint points

Images