| Submission name | Pyrender | ||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|
| Submission time (UTC) | May 13, 2025, 1:14 p.m. | ||||||||||
| User | nvnguyen | ||||||||||
| Task | Model-based 6D localization of unseen objects | ||||||||||
| Dataset | LM-O | ||||||||||
| Description | |||||||||||
| Evaluation scores |
|
| User | nvnguyen |
|---|---|
| Publication | https://arxiv.org/abs/2506.07155 |
| Implementation | https://github.com/facebookresearch/gotrack |
| Training image modalities | RGB-D |
| Test image modalities | RGB |
| Description | FoundPose (ECCV 2024): https://arxiv.org/abs/2311.18809 GoTrack (CVPRW 2025) is an efficient and accurate CAD-based method for 6DoF pose refinement and tracking of unseen objects. Given a CAD model of an object, an RGB image with known intrinsics that shows the object in an unknown pose, and an initial object pose, GoTrack refines the object pose such as the 2D projection of the model aligns closely with the object’s appearance in the image. Additional details: We first use FoundPose to estimate the coarse object pose, and then refine it using GoTrack with 5 iterations. For the leaderboard, we report the results using our open-source code, which differs slightly from those in our paper due to the use of a different renderer. |
| Computer specifications | V100 |