| Submission name | |||||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Submission time (UTC) | Jan. 27, 2026, 2:11 p.m. | ||||||||||||||
| User | SEU_WYL | ||||||||||||||
| Task | Model-based 6D detection of unseen objects | ||||||||||||||
| Dataset | IC-BIN | ||||||||||||||
| Description | |||||||||||||||
| Evaluation scores |
|
| User | SEU_WYL |
|---|---|
| Publication | WAPR: A Foundation Model for Wide-Angle Refinement in Unseen Object Pose Estimation, ECCV 2026 |
| Implementation | https://github.com/WangYuLin-SEU/WAPR |
| Training image modalities | RGB-D |
| Test image modalities | RGB-D |
| Description | WAPR is applied on top of the method at https://bop.felk.cvut.cz/method_info/1063/ to further refine the pose, i.e., it refines only the top 1 pose. The refinement is run for five iterations, and FoundationPose ScoreNet then selects the best output from six poses, including one initial pose and five refined poses. WAPR: A Foundation Model for Wide-Angle Refinement in Unseen Object Pose Estimation, ECCV 2026 Wang, Yulin and Hu, Mengting and Li, Hongli and Zhou, Jianghao and Luo, Chen |
| Computer specifications |