mlsanityData to 7 Oct 2026

Optimising 2D Pose Representation: Improve Accuracy, Stability and Generalisability Within Unsupervised 2D-3D Human Pose Estimation

arXiv 2209.00618 ↗First posted 1 Sep 2022cs.CVcited by 0

Not on the map. Why?

Abstract (arXiv)

This paper addresses the problem of 2D pose representation during unsupervised 2D to 3D pose lifting to improve the accuracy, stability and generalisability of 3D human pose estimation (HPE) models. All unsupervised 2D-3D HPE approaches provide the entire 2D kinematic skeleton to a model during training. We argue that this is sub-optimal and disruptive as long-range correlations are induced between independent 2D key points and predicted 3D ordinates during training. To this end, we conduct the following study. With a maximum architecture capacity of 6 residual blocks, we evaluate the performance of 5 models which each represent a 2D pose differently during the adversarial unsupervised 2D-3D HPE process. Additionally, we show the correlations between 2D key points which are learned during the training process, highlighting the unintuitive correlations induced when an entire 2D pose is provided to a lifting model. Our results show that the most optimal representation of a 2D pose is that of two independent segments, the torso and legs, with no shared features between each lifting network. This approach decreased the average error by 20% on the Human3.6M dataset when compared to a model with a near identical parameter count trained on the entire 2D kinematic skeleton. Furthermore, due to the complex nature of adversarial learning, we show how this representation can also improve convergence during training allowing for an optimum result to be obtained more often.

Key facts. No citing papers in the map yet (data through 7 Oct 2026). Builds on 29 of its 38 references that are in the map. First posted on arXiv 1 Sep 2022.

Builds on

29 of 38 references.

References · 29

  1. 2014Adam: A Method for Stochastic Optimizationcited by 36,960
  2. 2017Improved Training of Wasserstein GANscited by 2,223
  3. 2018OpenPose: Realtime Multi-Person 2D Pose Estimation using Part Affinity Fieldscited by 1,161
  4. 2016Least Squares Generative Adversarial Networkscited by 1,000
  5. 2016Stacked Hourglass Networks for Human Pose Estimationcited by 913
  6. 2017End-to-end Recovery of Human Shape and Posecited by 883
  7. 2016Monocular 3D Human Pose Estimation In The Wild Using Improved CNN Supervisioncited by 493
  8. 2017Towards Principled Methods for Training Generative Adversarial Networkscited by 460
  9. 2017A simple yet effective baseline for 3d human pose estimationcited by 401
  10. 20183D human pose estimation in video with temporal convolutions and semi-supervised trainingcited by 401
  11. 2016Coarse-to-Fine Volumetric Prediction for Single-Image 3D Human Posecited by 220
  12. 2019Semantic Graph Convolutional Networks for 3D Human Pose Regressioncited by 204
17 more
  1. 2017Towards 3D Human Pose Estimation in the Wild: a Weakly-supervised Approachcited by 114
  2. 20163D Human Pose Estimation = 2D Pose Estimation + Matchingcited by 112
  3. 2018Ordinal Depth Supervision for 3D Human Pose Estimationcited by 107
  4. 2019Unsupervised 3D Pose Estimation with Geometric Self-Supervisioncited by 95
  5. 2020SRNet: Improving Generalization in 3D Human Pose Estimation with a Split-and-Recombine Approachcited by 94
  6. 20183D Human Pose Estimation in the Wild by Adversarial Learningcited by 86
  7. 2019RepNet: Weakly Supervised Training of an Adversarial Reprojection Network for 3D Human Pose Estimationcited by 85
  8. 2017Lifting from the Deep: Convolutional 3D Pose Estimation from a Single Imagecited by 73
  9. 20203D Human Pose Estimation using Spatio-Temporal Networks with Explicit Occlusion Trainingcited by 54
  10. 2018Can 3D Pose be Learned from 2D Projections Alone?cited by 44
  11. 2015Maximum-Margin Structured Learning with Deep Networks for 3D Human Pose Estimationcited by 37
  12. 2021ElePose: Unsupervised 3D Human Pose Estimation by Predicting Camera Elevation and Learning Normalizing Flows on 2D Posescited by 35
  13. 2020Kinematic-Structure-Preserved Representation for Unsupervised 3D Human Pose Estimationcited by 26
  14. 2018Unsupervised Adversarial Learning of 3D Human Pose from 2D Joint Locationscited by 18
  15. 20183D Human Pose Estimation with Relational Networkscited by 10
  16. 2017A Kinematic Chain Space for Monocular Motion Capturecited by 8
  17. 2018It's all Relative: Monocular 3D Human Pose Estimation from Weakly Supervised Datacited by 5

Built on by

Not cited yet.