Get object pose that is consistent with the input and is unique with C-3PO

Get object pose that is consistent with the input and is unique with C-3PO

Certifiable 3D Object Pose Estimation: Foundations, Learning Models, and Self-Training
arXiv paper abstract https://arxiv.org/abs/2206.11215
arXiv PDF paper https://arxiv.org/pdf/2206.11215.pdf
GitHub https://github.com/mit-spark/c-3po

… consider a certifiable object pose estimation problem … given a partial point cloud of an object — not only estimate the object pose, but also to provide a certificate of correctness for the resulting estimate.

… implementing two certificates: (i) a certificate of observable correctness, that asserts if the model output is consistent with the input data and prior information, (ii) a certificate of non-degeneracy, that asserts whether the input data is sufficient to compute a unique estimate.

… propose C-3PO, a semantic-keypoint-based pose estimation model, augmented with the two certificates, to solve the certifiable pose estimation problem.

C-3PO also includes a keypoint corrector … that can correct large detection errors (e.g. due to the sim-to-real gap).

… third contribution is a novel self-supervised training approach that uses … certificate of observable correctness to provide the supervisory signal to C-3PO during training.

… experiments show that (i) standard semantic-keypoint-based methods outperform more recent alternatives, (ii) C-3PO … significantly outperforms all the baselines, and (iii) C-3PO’s certificates are able to discern correct pose estimates.

Stay up to date. Subscribe to my posts https://morrislee1234.wixsite.com/website/contact
Web site with my other posts by category https://morrislee1234.wixsite.com/website

LinkedIn https://www.linkedin.com/in/morris-lee-47877b7b

Photo by Maria Lupan on Unsplash