CaltechAUTHORS
  A Caltech Library Service

It's all Relative: Monocular 3D Human Pose Estimation from Weakly Supervised Data

Ronchi, Matteo Ruggero and Mac Aodha, Oisin and Eng, Robert and Perona, Pietro (2018) It's all Relative: Monocular 3D Human Pose Estimation from Weakly Supervised Data. . (Unpublished) http://resolver.caltech.edu/CaltechAUTHORS:20180613-133929571

[img] PDF - Submitted Version
See Usage Policy.

4Mb

Use this Persistent URL to link to this item: http://resolver.caltech.edu/CaltechAUTHORS:20180613-133929571

Abstract

We address the problem of 3D human pose estimation from 2D input images using only weakly supervised training data. Despite showing considerable success for 2D pose estimation, the application of supervised machine learning to 3D pose estimation in real world images is currently hampered by the lack of varied training images with associated 3D poses. Existing 3D pose estimation algorithms train on data that has either been collected in carefully controlled studio settings or has been generated synthetically. Instead, we take a different approach, and propose a 3D human pose estimation algorithm that only requires relative estimates of depth at training time. Such training signal, although noisy, can be easily collected from crowd annotators, and is of sufficient quality for enabling successful training and evaluation of 3D pose. Our results are competitive with fully supervised regression based approaches on the Human3.6M dataset, despite using significantly weaker training data. Our proposed approach opens the door to using existing widespread 2D datasets for 3D pose estimation by allowing fine-tuning with noisy relative constraints, resulting in more accurate 3D poses.


Item Type:Report or Paper (Discussion Paper)
Related URLs:
URLURL TypeDescription
http://arxiv.org/abs/1805.06880arXivDiscussion Paper
ORCID:
AuthorORCID
Perona, Pietro0000-0002-7583-5809
Additional Information:© 2018. The copyright of this document resides with its authors. It may be distributed unchanged freely in print or electronic forms. We would like to thank Google for their gift to the Visipedia project and Amazon Web Services (AWS) for Research Credits.
Funders:
Funding AgencyGrant Number
GoogleUNSPECIFIED
Amazon Web ServicesUNSPECIFIED
Record Number:CaltechAUTHORS:20180613-133929571
Persistent URL:http://resolver.caltech.edu/CaltechAUTHORS:20180613-133929571
Usage Policy:No commercial reproduction, distribution, display or performance rights in this work are provided.
ID Code:87073
Collection:CaltechAUTHORS
Deposited By: Caroline Murphy
Deposited On:13 Jun 2018 20:58
Last Modified:13 Jun 2018 20:58

Repository Staff Only: item control page