Human pose estimation via convolutional part heatmap regression
- Submitting institution
-
University of Nottingham, The
- Unit of assessment
- 11 - Computer Science and Informatics
- Output identifier
- 1330520
- Type
- E - Conference contribution
- DOI
-
10.1007/978-3-319-46478-7_44
- Title of conference / published proceedings
- Computer Vision – ECCV 2016
- First page
- 717
- Volume
- 9911
- Issue
- -
- ISSN
- -
- Open access status
- -
- Month of publication
- September
- Year of publication
- 2016
- URL
-
-
- Supplementary information
-
-
- Request cross-referral to
- -
- Output has been delayed by COVID-19
- No
- COVID-19 affected output statement
- -
- Forensic science
- No
- Criminology
- No
- Interdisciplinary
- No
- Number of additional authors
-
1
- Research group(s)
-
-
- Citation count
- 133
- Proposed double-weighted
- No
- Reserve for an output with double weighting
- No
- Additional information
- Applications of human pose detection span many areas, from entertainment (such as motion capture) to healthcare (such as falling patients). Prior approaches typically opt for regression in Euclidean space. This paper describes the first human pose network to use 2D (pixel to pixel) heatmap regression to identify joint locations. The technique also employs part detection before point regression, reinforcing possible poses. Published at a high- ranking conference with an average acceptance rate of 27%, the proposed method offers state of the art performance and has been experimentally compared with 10 similar works.
- Author contribution statement
- -
- Non-English
- No
- English abstract
- -
