DiveNet: Dive Action Localization and Physical Pose Parameter Extraction for High Performance Training
The tremendous progress of deep convolution neural networks has shown promising results on the classification of various sports activities. However, the accurate localization of a particular sports event or activity in a continuous video stream is still a challenging problem. The accurate detection...
| Main Authors: | , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
IEEE
2023-01-01
|
| Series: | IEEE Access |
| Subjects: | |
| Online Access: | https://ieeexplore.ieee.org/document/10097490/ |
| _version_ | 1827962807040081920 |
|---|---|
| author | Pramod Murthy Bertram Taetz Arpit Lekhra Didier Stricker |
| author_facet | Pramod Murthy Bertram Taetz Arpit Lekhra Didier Stricker |
| author_sort | Pramod Murthy |
| collection | DOAJ |
| description | The tremendous progress of deep convolution neural networks has shown promising results on the classification of various sports activities. However, the accurate localization of a particular sports event or activity in a continuous video stream is still a challenging problem. The accurate detection of sports actions enables the comparison of different performances, objectively. In this work, we propose the DiveNet action localization module to detect the springboard diving sports action in an unconstrained environment. We used Temporal Convolution Network (TCN) over a backbone feature extractor to localize diving actions, with low latency. We estimate the divers center of mass (COM) trajectory and the peak dive height using the temporal demarcations provided by the action localization step via the projectile motion formula. In addition, we train a DiveNet pose regression network, which extends the Unipose architecture with direct physical parameter estimation, i.e COM and 2D joint keypoints. We propose a new homography computation method between the diving motion plane and the image-view for each dive. This enables the representation of physical parameters in metric scale, without any calibration. We release the first publicly available diving sports video dataset, recorded at 60 Hz with a static camera setup for different springboard heights. DiveNet action localization achieves an accuracy of 95% with a single frame latency (< 25 ms). The DiveNet pose regression model shows competitive results around 70% PCK on different diving pose datasets. We achieve COM accuracy of 6 pixels, dive peak height sensitivity of 20 cm and mean joint angle errors around 10 degrees. |
| first_indexed | 2024-04-09T16:48:43Z |
| format | Article |
| id | doaj.art-1db37321f57142a289c6182f4000090c |
| institution | Directory Open Access Journal |
| issn | 2169-3536 |
| language | English |
| last_indexed | 2024-04-09T16:48:43Z |
| publishDate | 2023-01-01 |
| publisher | IEEE |
| record_format | Article |
| series | IEEE Access |
| spelling | doaj.art-1db37321f57142a289c6182f4000090c2023-04-21T23:00:07ZengIEEEIEEE Access2169-35362023-01-0111377493776710.1109/ACCESS.2023.326559510097490DiveNet: Dive Action Localization and Physical Pose Parameter Extraction for High Performance TrainingPramod Murthy0https://orcid.org/0000-0002-8016-8537Bertram Taetz1https://orcid.org/0000-0001-9921-4874Arpit Lekhra2https://orcid.org/0009-0002-2001-7400Didier Stricker3Department of Augmented Vision, Rheinland-Pfälzische Technische Universität Kaiserslautern-Landau (RPTU), Kaiserslautern, GermanyDepartment of Augmented Vision, German Research Center for Artificial Intelligence (DFKI), Kaiserslautern, GermanyDepartment of Augmented Vision, Rheinland-Pfälzische Technische Universität Kaiserslautern-Landau (RPTU), Kaiserslautern, GermanyDepartment of Augmented Vision, Rheinland-Pfälzische Technische Universität Kaiserslautern-Landau (RPTU), Kaiserslautern, GermanyThe tremendous progress of deep convolution neural networks has shown promising results on the classification of various sports activities. However, the accurate localization of a particular sports event or activity in a continuous video stream is still a challenging problem. The accurate detection of sports actions enables the comparison of different performances, objectively. In this work, we propose the DiveNet action localization module to detect the springboard diving sports action in an unconstrained environment. We used Temporal Convolution Network (TCN) over a backbone feature extractor to localize diving actions, with low latency. We estimate the divers center of mass (COM) trajectory and the peak dive height using the temporal demarcations provided by the action localization step via the projectile motion formula. In addition, we train a DiveNet pose regression network, which extends the Unipose architecture with direct physical parameter estimation, i.e COM and 2D joint keypoints. We propose a new homography computation method between the diving motion plane and the image-view for each dive. This enables the representation of physical parameters in metric scale, without any calibration. We release the first publicly available diving sports video dataset, recorded at 60 Hz with a static camera setup for different springboard heights. DiveNet action localization achieves an accuracy of 95% with a single frame latency (< 25 ms). The DiveNet pose regression model shows competitive results around 70% PCK on different diving pose datasets. We achieve COM accuracy of 6 pixels, dive peak height sensitivity of 20 cm and mean joint angle errors around 10 degrees.https://ieeexplore.ieee.org/document/10097490/Deep learningdiving sports poseaction localizationsports analytics |
| spellingShingle | Pramod Murthy Bertram Taetz Arpit Lekhra Didier Stricker DiveNet: Dive Action Localization and Physical Pose Parameter Extraction for High Performance Training IEEE Access Deep learning diving sports pose action localization sports analytics |
| title | DiveNet: Dive Action Localization and Physical Pose Parameter Extraction for High Performance Training |
| title_full | DiveNet: Dive Action Localization and Physical Pose Parameter Extraction for High Performance Training |
| title_fullStr | DiveNet: Dive Action Localization and Physical Pose Parameter Extraction for High Performance Training |
| title_full_unstemmed | DiveNet: Dive Action Localization and Physical Pose Parameter Extraction for High Performance Training |
| title_short | DiveNet: Dive Action Localization and Physical Pose Parameter Extraction for High Performance Training |
| title_sort | divenet dive action localization and physical pose parameter extraction for high performance training |
| topic | Deep learning diving sports pose action localization sports analytics |
| url | https://ieeexplore.ieee.org/document/10097490/ |
| work_keys_str_mv | AT pramodmurthy divenetdiveactionlocalizationandphysicalposeparameterextractionforhighperformancetraining AT bertramtaetz divenetdiveactionlocalizationandphysicalposeparameterextractionforhighperformancetraining AT arpitlekhra divenetdiveactionlocalizationandphysicalposeparameterextractionforhighperformancetraining AT didierstricker divenetdiveactionlocalizationandphysicalposeparameterextractionforhighperformancetraining |