File Download

There are no files associated with this item.

  Links for fulltext
     (May Require Subscription)
Supplementary

Conference Paper: Mapping in a Cycle: Sinkhorn Regularized Unsupervised Learning for Point Cloud Shapes

TitleMapping in a Cycle: Sinkhorn Regularized Unsupervised Learning for Point Cloud Shapes
Authors
KeywordsPoint cloud
Unsupervised learning
Dense correspondence
Cycle-consistency
Issue Date2020
PublisherSpringer.
Citation
Proceedings of the 16th European Conference on Computer Vision (ECCV), Online, Glasgow, UK, 23-28 August 2020, pt X, p. 455-472 How to Cite?
AbstractWe propose an unsupervised learning framework with the pretext task of finding dense correspondences between point cloud shapes from the same category based on the cycle-consistency formulation. In order to learn discriminative pointwise features from point cloud data, we incorporate in the formulation a regularization term based on Sinkhorn normalization to enhance the learned pointwise mappings to be as bijective as possible. Besides, a random rigid transform of the source shape is introduced to form a triplet cycle to improve the model’s robustness against perturbations. Comprehensive experiments demonstrate that the learned pointwise features through our framework benefits various point cloud analysis tasks, e.g. partial shape registration and keypoint transfer. We also show that the learned pointwise features can be leveraged by supervised methods to improve the part segmentation performance with either the full training dataset or just a small portion of it.
Persistent Identifierhttp://hdl.handle.net/10722/293456
ISBN
Series/Report no.Lecture Notes in Computer Science (LNCS), v. 12355

 

DC FieldValueLanguage
dc.contributor.authorYang, L-
dc.contributor.authorLiu, W-
dc.contributor.authorCui, Z-
dc.contributor.authorChen, N-
dc.contributor.authorWang, WP-
dc.date.accessioned2020-11-23T08:17:01Z-
dc.date.available2020-11-23T08:17:01Z-
dc.date.issued2020-
dc.identifier.citationProceedings of the 16th European Conference on Computer Vision (ECCV), Online, Glasgow, UK, 23-28 August 2020, pt X, p. 455-472-
dc.identifier.isbn9783030586065-
dc.identifier.urihttp://hdl.handle.net/10722/293456-
dc.description.abstractWe propose an unsupervised learning framework with the pretext task of finding dense correspondences between point cloud shapes from the same category based on the cycle-consistency formulation. In order to learn discriminative pointwise features from point cloud data, we incorporate in the formulation a regularization term based on Sinkhorn normalization to enhance the learned pointwise mappings to be as bijective as possible. Besides, a random rigid transform of the source shape is introduced to form a triplet cycle to improve the model’s robustness against perturbations. Comprehensive experiments demonstrate that the learned pointwise features through our framework benefits various point cloud analysis tasks, e.g. partial shape registration and keypoint transfer. We also show that the learned pointwise features can be leveraged by supervised methods to improve the part segmentation performance with either the full training dataset or just a small portion of it.-
dc.languageeng-
dc.publisherSpringer.-
dc.relation.ispartofEuropean Conference on Computer Vision (ECCV 2020)-
dc.relation.ispartofseriesLecture Notes in Computer Science (LNCS), v. 12355-
dc.subjectPoint cloud-
dc.subjectUnsupervised learning-
dc.subjectDense correspondence-
dc.subjectCycle-consistency-
dc.titleMapping in a Cycle: Sinkhorn Regularized Unsupervised Learning for Point Cloud Shapes-
dc.typeConference_Paper-
dc.identifier.emailYang, L: lyang125@hku.hk-
dc.identifier.emailWang, WP: wenping@cs.hku.hk-
dc.identifier.authorityWang, WP=rp00186-
dc.description.naturelink_to_subscribed_fulltext-
dc.identifier.doi10.1007/978-3-030-58607-2_27-
dc.identifier.hkuros318909-
dc.identifier.volumept X-
dc.identifier.spage455-
dc.identifier.epage472-
dc.publisher.placeCham-

Export via OAI-PMH Interface in XML Formats


OR


Export to Other Non-XML Formats