A 3D Keypoints Voting Network for 6DoF Pose Estimation in Indoor Scene
This paper addresses the problem of instance-level 6DoF pose estimation from a single RGBD image in an indoor scene. Many recent works have shown that a two-stage network, which first detects the keypoints and then regresses the keypoints for 6d pose estimation, achieves remarkable performance. Howe...
| Main Authors: | , , , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
MDPI AG
2021-10-01
|
| Series: | Machines |
| Subjects: | |
| Online Access: | https://www.mdpi.com/2075-1702/9/10/230 |
| _version_ | 1827679257688539136 |
|---|---|
| author | Huikai Liu Gaorui Liu Yue Zhang Linjian Lei Hui Xie Yan Li Shengli Sun |
| author_facet | Huikai Liu Gaorui Liu Yue Zhang Linjian Lei Hui Xie Yan Li Shengli Sun |
| author_sort | Huikai Liu |
| collection | DOAJ |
| description | This paper addresses the problem of instance-level 6DoF pose estimation from a single RGBD image in an indoor scene. Many recent works have shown that a two-stage network, which first detects the keypoints and then regresses the keypoints for 6d pose estimation, achieves remarkable performance. However, the previous methods concern little about channel-wise attention and the keypoints are not selected by comprehensive use of RGBD information, which limits the performance of the network. To enhance RGB feature representation ability, a modular Split-Attention block that enables attention across feature-map groups is proposed. In addition, by combining the Oriented FAST and Rotated BRIEF (ORB) keypoints and the Farthest Point Sample (FPS) algorithm, a simple but effective keypoint selection method named ORB-FPS is presented to avoid the keypoints appear on the non-salient regions. The proposed algorithm is tested on the Linemod and the YCB-Video dataset, the experimental results demonstrate that our method outperforms the current approaches, achieves ADD(S) accuracy of 94.5% on the Linemod dataset and 91.4% on the YCB-Video dataset. |
| first_indexed | 2024-03-10T06:26:49Z |
| format | Article |
| id | doaj.art-51912857e6004a448625cdaaa33bfb00 |
| institution | Directory Open Access Journal |
| issn | 2075-1702 |
| language | English |
| last_indexed | 2024-03-10T06:26:49Z |
| publishDate | 2021-10-01 |
| publisher | MDPI AG |
| record_format | Article |
| series | Machines |
| spelling | doaj.art-51912857e6004a448625cdaaa33bfb002023-11-22T18:54:26ZengMDPI AGMachines2075-17022021-10-0191023010.3390/machines9100230A 3D Keypoints Voting Network for 6DoF Pose Estimation in Indoor SceneHuikai Liu0Gaorui Liu1Yue Zhang2Linjian Lei3Hui Xie4Yan Li5Shengli Sun6Shanghai Institute of Technical Physics, Chinese Academy of Sciences, Shanghai 200083, ChinaShanghai Institute of Technical Physics, Chinese Academy of Sciences, Shanghai 200083, ChinaShanghai Institute of Technical Physics, Chinese Academy of Sciences, Shanghai 200083, ChinaShanghai Institute of Technical Physics, Chinese Academy of Sciences, Shanghai 200083, ChinaShanghai Institute of Technical Physics, Chinese Academy of Sciences, Shanghai 200083, ChinaShanghai Institute of Technical Physics, Chinese Academy of Sciences, Shanghai 200083, ChinaShanghai Institute of Technical Physics, Chinese Academy of Sciences, Shanghai 200083, ChinaThis paper addresses the problem of instance-level 6DoF pose estimation from a single RGBD image in an indoor scene. Many recent works have shown that a two-stage network, which first detects the keypoints and then regresses the keypoints for 6d pose estimation, achieves remarkable performance. However, the previous methods concern little about channel-wise attention and the keypoints are not selected by comprehensive use of RGBD information, which limits the performance of the network. To enhance RGB feature representation ability, a modular Split-Attention block that enables attention across feature-map groups is proposed. In addition, by combining the Oriented FAST and Rotated BRIEF (ORB) keypoints and the Farthest Point Sample (FPS) algorithm, a simple but effective keypoint selection method named ORB-FPS is presented to avoid the keypoints appear on the non-salient regions. The proposed algorithm is tested on the Linemod and the YCB-Video dataset, the experimental results demonstrate that our method outperforms the current approaches, achieves ADD(S) accuracy of 94.5% on the Linemod dataset and 91.4% on the YCB-Video dataset.https://www.mdpi.com/2075-1702/9/10/2306DoF pose estimationsplit-channel attentionORB-FPS keypoint |
| spellingShingle | Huikai Liu Gaorui Liu Yue Zhang Linjian Lei Hui Xie Yan Li Shengli Sun A 3D Keypoints Voting Network for 6DoF Pose Estimation in Indoor Scene Machines 6DoF pose estimation split-channel attention ORB-FPS keypoint |
| title | A 3D Keypoints Voting Network for 6DoF Pose Estimation in Indoor Scene |
| title_full | A 3D Keypoints Voting Network for 6DoF Pose Estimation in Indoor Scene |
| title_fullStr | A 3D Keypoints Voting Network for 6DoF Pose Estimation in Indoor Scene |
| title_full_unstemmed | A 3D Keypoints Voting Network for 6DoF Pose Estimation in Indoor Scene |
| title_short | A 3D Keypoints Voting Network for 6DoF Pose Estimation in Indoor Scene |
| title_sort | 3d keypoints voting network for 6dof pose estimation in indoor scene |
| topic | 6DoF pose estimation split-channel attention ORB-FPS keypoint |
| url | https://www.mdpi.com/2075-1702/9/10/230 |
| work_keys_str_mv | AT huikailiu a3dkeypointsvotingnetworkfor6dofposeestimationinindoorscene AT gaoruiliu a3dkeypointsvotingnetworkfor6dofposeestimationinindoorscene AT yuezhang a3dkeypointsvotingnetworkfor6dofposeestimationinindoorscene AT linjianlei a3dkeypointsvotingnetworkfor6dofposeestimationinindoorscene AT huixie a3dkeypointsvotingnetworkfor6dofposeestimationinindoorscene AT yanli a3dkeypointsvotingnetworkfor6dofposeestimationinindoorscene AT shenglisun a3dkeypointsvotingnetworkfor6dofposeestimationinindoorscene AT huikailiu 3dkeypointsvotingnetworkfor6dofposeestimationinindoorscene AT gaoruiliu 3dkeypointsvotingnetworkfor6dofposeestimationinindoorscene AT yuezhang 3dkeypointsvotingnetworkfor6dofposeestimationinindoorscene AT linjianlei 3dkeypointsvotingnetworkfor6dofposeestimationinindoorscene AT huixie 3dkeypointsvotingnetworkfor6dofposeestimationinindoorscene AT yanli 3dkeypointsvotingnetworkfor6dofposeestimationinindoorscene AT shenglisun 3dkeypointsvotingnetworkfor6dofposeestimationinindoorscene |