File Download

There are no files associated with this item.

  Links for fulltext
     (May Require Subscription)
Supplementary

Conference Paper: DISP6D: Disentangled Implicit Shape and Pose Learning for Scalable 6D Pose Estimation

TitleDISP6D: Disentangled Implicit Shape and Pose Learning for Scalable 6D Pose Estimation
Authors
Keywords6D pose estimation
Disentanglement
Re-entanglement
Scalability
Sim-to-real
Symmetry ambiguity
Issue Date6-Nov-2022
Abstract

Scalable 6D pose estimation for rigid objects from RGB images aims at handling multiple objects and generalizing to novel objects. Building on a well-known auto-encoding framework to cope with object symmetry and the lack of labeled training data, we achieve scalability by disentangling the latent representation of auto-encoder into shape and pose sub-spaces. The latent shape space models the similarity of different objects through contrastive metric learning, and the latent pose code is compared with canonical rotations for rotation retrieval. Because different object symmetries induce inconsistent latent pose spaces, we re-entangle the shape representation with canonical rotations to generate shape-dependent pose codebooks for rotation retrieval. We show state-of-the-art performance on two benchmarks containing textureless CAD objects without category and daily objects with categories respectively, and further demonstrate improved scalability by extending to a more challenging setting of daily objects across categories.


Persistent Identifierhttp://hdl.handle.net/10722/333852
ISI Accession Number ID

 

DC FieldValueLanguage
dc.contributor.authorWen, YL-
dc.contributor.authorLi, XY-
dc.contributor.authorPan, H-
dc.contributor.authorYang, L-
dc.contributor.authorWang, Z-
dc.contributor.authorKomura, T-
dc.contributor.authorWang, WP-
dc.date.accessioned2023-10-06T08:39:37Z-
dc.date.available2023-10-06T08:39:37Z-
dc.date.issued2022-11-06-
dc.identifier.urihttp://hdl.handle.net/10722/333852-
dc.description.abstract<p>Scalable 6D pose estimation for rigid objects from RGB images aims at handling multiple objects and generalizing to novel objects. Building on a well-known auto-encoding framework to cope with object symmetry and the lack of labeled training data, we achieve scalability by disentangling the latent representation of auto-encoder into shape and pose sub-spaces. The latent shape space models the similarity of different objects through contrastive metric learning, and the latent pose code is compared with canonical rotations for rotation retrieval. Because different object symmetries induce inconsistent latent pose spaces, we re-entangle the shape representation with canonical rotations to generate shape-dependent pose codebooks for rotation retrieval. We show state-of-the-art performance on two benchmarks containing textureless CAD objects without category and daily objects with categories respectively, and further demonstrate improved scalability by extending to a more challenging setting of daily objects across categories.</p>-
dc.languageeng-
dc.relation.ispartof17th European Conference on Computer Vision, ECCV 2022 (23/10/2022-27/10/2022, Tel Aviv, Israel)-
dc.subject6D pose estimation-
dc.subjectDisentanglement-
dc.subjectRe-entanglement-
dc.subjectScalability-
dc.subjectSim-to-real-
dc.subjectSymmetry ambiguity-
dc.titleDISP6D: Disentangled Implicit Shape and Pose Learning for Scalable 6D Pose Estimation-
dc.typeConference_Paper-
dc.identifier.doi10.1007/978-3-031-20077-9_24-
dc.identifier.scopuseid_2-s2.0-85142754501-
dc.identifier.volume13669 LNCS-
dc.identifier.spage404-
dc.identifier.epage421-
dc.identifier.isiWOS:000897132300024-

Export via OAI-PMH Interface in XML Formats


OR


Export to Other Non-XML Formats