3D Human Pose Estimation With Spatial Structure Information

Abstract

Estimating 3D human poses from 2D poses is a challenging problem due to joints selfocclusion, weak generalization, and inherent ambiguity of recovering depth. Actually, there exists spatial structure dependence on human body key points which can be used to alleviate the problem of joints selfocclusion. Therefore, we represent human pose as a directed graph and propose a network implemented with graph convolution to predict 3D poses from the given 2D poses. In the digraph, we determine the connection weight of each edge according to the error distribution of joints estimation. This makes our model robust to noise. By optimizing coarse 3D estimation and adversarial learning, our algorithm can successfully improve the accuracy of estimation and relieve the ambiguity of mapping. Through testing on Human 3.6M and MPI-INF-3DHP datasets, we achieve excellent quantitative performance. More importantly, our algorithm also has a superior generalization to outdoor dataset MPII by the pre-training process.INDEX TERMS 3D human poses, graph convolutional networks, adversarial learning, geometric priors, gradient vanish, in-the-wild scenes.

Accuracy Assessment of Joint Angles Estimated from 2D and 3D Camera Measurements

“…This makes it hard to compare it directly to many other camera based methods, because the large majority uses position error metrics instead of angles. Even though joint angle errors are the most indicative metric for ergonomics, lost methods use MPJPE ( [13][14][15][22][23][24][25]), PCP ([16,17,20,24,26-28]) or PCK ([13,15,22,28]).…”

“…In the field of 3D human pose estimation there is much research on single view 3D estimation [13][14][15]. They have the advantage that they can be used in many applications, as only a simple camera is needed.…”

A Novel Approach of Intelligent Computing for Multiperson Pose Estimation with Deep High Spatial Resolution and Multiscale Features

“…Recently, the human pose estimation methods based on Convolutional Neural Networks (CNNs) have achieved a great breakthrough [1][2][3][4][5][6], since CNNs have the powerful ability to learn rich convolutional feature representations [7]. For example, for singleperson pose estimation, the state-of-the-art models have improved the performance from less than 50% PCKh@0.5 to more than 90% PCKh@0.5 [8][9][10][11][12] on the MPII benchmark [13]. However, multiperson pose estimation still faces two main challenges:…”