Evaluating current state of monocular 3D pose models for golf

Christian Keilstrup Ingwersen, Janus Nørtoft Jensen, Morten Rieger Hannemose, Anders Bjorholm Dahl

Research output: Chapter in Book/Report/Conference proceedingArticle in proceedingsResearchpeer-review

305 Downloads (Pure)

Abstract

Monocular 3D human pose estimation has reached an impressive performance. State-of-the-art models predict joint locations that can be accurately reprojected back into the image, resulting in visually convincing detections. However, our aim is to use the predicted poses in a domain with high- frequency movements, that is, for video of athletes performing golf swings. Our investigation is based on accurate marker-based motion capture data. Also, for our data, the predicted 3D joint locations look convincing when we reproject them into the image. However, by quantitatively com- paring the results with the motion capture data, we see significant model errors that are too erroneous to be used for any kinematic analysis of the movements. Thus we conclude that the current models cannot be used out of the box for advanced golf analytics.
Original languageEnglish
Title of host publicationProceedings of the Northern Lights Deep Learning Workshop 2023
Number of pages10
Volume4
PublisherSeptentrio Academic Publishing
Publication date2023
DOIs
Publication statusPublished - 2023
Event Northern Lights Deep Learning Workshop 2023 - Tromsø, Norway
Duration: 10 Jan 202312 Jan 2023

Workshop

Workshop Northern Lights Deep Learning Workshop 2023
Country/TerritoryNorway
CityTromsø
Period10/01/202312/01/2023

Fingerprint

Dive into the research topics of 'Evaluating current state of monocular 3D pose models for golf'. Together they form a unique fingerprint.

Cite this