Sim2real Cattle Pose Prediction in 3D pointclouds

Okour, M; Guenot-Falque, R; Vidal Calleja, T; Alempijevic, A

Sim2real Cattle Pose Prediction in 3D pointclouds

Okour, M Guenot-Falque, R Vidal Calleja, T Alempijevic, A

Permalink

Publisher:: ARAA
Publication Type:: Conference Proceeding
Citation:: https://ssl.linklings.net/conferences/acra/acra2022_proceedings/views/by_auth.html, 2022, pp. 1-8
Issue Date:: 2022-12-06

Open Access

Copyright Clearance Process

Recently Added
In Progress
Open Access

This item is open access.

Adobe PDF

Download Accepted versionAdobe PDF (4.17 MB)

View statistics

Full metadata record

Field	Value	Language
dc.contributor.author	Okour, M
dc.contributor.author	Guenot-Falque, R
dc.contributor.author	Vidal Calleja, T
dc.contributor.author	Alempijevic, A https://orcid.org/0000-0002-1769-8041
dc.date	2022-12-06
dc.date.accessioned	2022-12-09T07:39:16Z
dc.date.available	2022-11-03
dc.date.available	2022-12-09T07:39:16Z
dc.date.issued	2022-12-06
dc.identifier.citation	https://ssl.linklings.net/conferences/acra/acra2022_proceedings/views/by_auth.html, 2022, pp. 1-8
dc.identifier.uri	http://hdl.handle.net/10453/164272
dc.description.abstract	Cattle's body shape and joint articulation carry significant information about their well-being. Building a large dataset of any animals' 3D scans is a challenging task. However, such a dataset is required for training deep learning algorithms for 3D body pose estimation. In this work, we investigate how such a dataset can be constructed for cattle from a single 3D model animated by a digital artist. Further, we reduce the sim2real gap between the virtual dataset and real scans of animals by augmenting the shape of the 3D model to cover the range of possible body shapes. The generated dataset is tested on semantic keypoints detection with an encoder-decoder architecture.
dc.language	en
dc.publisher	ARAA
dc.relation	Food Agility CRC Limited
dc.relation.ispartof	https://ssl.linklings.net/conferences/acra/acra2022_proceedings/views/by_auth.html
dc.relation.ispartof	Australasian Conference on Robotics and Automation
dc.rights	info:eu-repo/semantics/openAccess
dc.title	Sim2real Cattle Pose Prediction in 3D pointclouds
dc.type	Conference Proceeding
utslib.location.activity	Brisbane, Australia
pubs.organisational-group	/University of Technology Sydney
pubs.organisational-group	/University of Technology Sydney/Faculty of Engineering and Information Technology
pubs.organisational-group	/University of Technology Sydney/Faculty of Engineering and Information Technology/School of Mechanical and Mechatronic Engineering
utslib.copyright.status	open_access	*
pubs.consider-herdc	false
dc.date.updated	2022-12-09T07:39:13Z
pubs.finish-date	2022-12-08
pubs.publication-status	Published online
pubs.start-date	2022-12-06

Abstract:

Cattle's body shape and joint articulation carry significant information about their well-being. Building a large dataset of any animals' 3D scans is a challenging task. However, such a dataset is required for training deep learning algorithms for 3D body pose estimation. In this work, we investigate how such a dataset can be constructed for cattle from a single 3D model animated by a digital artist. Further, we reduce the sim2real gap between the virtual dataset and real scans of animals by augmenting the shape of the 3D model to cover the range of possible body shapes. The generated dataset is tested on semantic keypoints detection with an encoder-decoder architecture.

Please use this identifier to cite or link to this item:

http://hdl.handle.net/10453/164272