3D action recognition from novel viewpoints
- Submitting institution
-
The University of Lancaster
- Unit of assessment
- 11 - Computer Science and Informatics
- Output identifier
- 249321400
- Type
- E - Conference contribution
- DOI
-
10.1109/CVPR.2016.167
- Title of conference / published proceedings
- 2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR)
- First page
- 1506
- Volume
- -
- Issue
- -
- ISSN
- -
- Open access status
- -
- Month of publication
- June
- 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)
-
B - Data Science
- Citation count
- 50
- Proposed double-weighted
- No
- Reserve for an output with double weighting
- No
- Additional information
- This paper presents a novel view-invariant deep learning based human activity recognition method from RGB videos. It addresses the issue of changes in the image pixel trajectories in images across different camera viewpoints by training a single deep neural network, which transfers pixel trajectories from multiple views into a canonical view. This paper is significant as it does not require action labels during learning, it uses a single model for all actions observed from all viewpoints and adding a new action class does not require retraining the model. The work was funded by an ARC discovery grant, Australia (AU$293K).
- Author contribution statement
- -
- Non-English
- No
- English abstract
- -
