Abstract
In this paper, we propose a new end-to-end articulated structure-aware network to regress 3D joint coordinates from the given 2D joint detections. The proposed method is capable of dealing with hard joints well that usually fail existing methods. Specifically, our framework cascades a refinement network with a basic network for two types of joints, and employs a attention module to simulate a camera projection model. In addition, we propose to use a random enhancement module to intensify the constraints between joints. Experimental results on the Human3.6M and HumanEva databases demonstrate the effectiveness and flexibility of the proposed network, and errors of hard joints and bone lengths are significantly reduced, compared with state-of-the-art approaches. © 2019 Z. Tang, X. Zhang & J. Hou.
| Original language | English |
|---|---|
| Title of host publication | Proceedings of The Eleventh Asian Conference on Machine Learning |
| Editors | Wee Sun Lee, Taiji Suzuki |
| Pages | 48-63 |
| Publication status | Published - Nov 2019 |
| Event | 11th Asian Conference on Machine Learning (ACML 2019) - WINC Aichi, Nagoya, Japan Duration: 17 Nov 2019 → 19 Nov 2019 https://www.acml-conf.org/2019/ |
Publication series
| Name | Proceedings of Machine Learning Research |
|---|---|
| Volume | 101 |
| ISSN (Print) | 2640-3498 |
Conference
| Conference | 11th Asian Conference on Machine Learning (ACML 2019) |
|---|---|
| Abbreviated title | ACML2019 |
| Country/Territory | Japan |
| City | Nagoya |
| Period | 17/11/19 → 19/11/19 |
| Internet address |
Research Keywords
- 3D human pose estimation
- Articulated structure-aware network
- Attention module
- Random enhancement module