Self-supervised learning for single view depth and surface normal estimation

Simple item page
dc.contributor.authorZhan, H.
dc.contributor.authorWeerasekera, C.S.
dc.contributor.authorGarg, R.
dc.contributor.authorReid, I.
dc.contributor.conferenceIEEE International Conference on Robotics and Automation (ICRA) (20 May 2019 - 24 May 2019 : Montreal, Canada)
dc.contributor.editorHoward, A.
dc.contributor.editorAlthoefer, K.
dc.contributor.editorArai, F.
dc.contributor.editorArrichiello, F.
dc.contributor.editorCaputo, B.
dc.contributor.editorCastellanos, J.
dc.contributor.editorHauser, K.
dc.contributor.editorIsler, V.
dc.contributor.editorKim, J.
dc.contributor.editorLiu, H.
dc.contributor.editorOh, P.
dc.contributor.editorSantos, V.
dc.contributor.editorScaramuzza, D.
dc.contributor.editorUde, A.
dc.contributor.editorVoyles, R.
dc.contributor.editorYamane, K.
dc.contributor.editorOkamura, A.
dc.date.issued2019
dc.description.abstractIn this work we present a self-supervised learning framework to simultaneously train two Convolutional Neural Networks (CNNs) to predict depth and surface normals from a single image. In contrast to most existing frameworks which represent outdoor scenes as fronto-parallel planes at piecewise smooth depth, we propose to predict depth with surface orientation while assuming that natural scenes have piece-wise smooth normals. We show that a simple depth-normal consistency as a soft-constraint on the predictions is sufficient and effective for training both these networks simultaneously. The trained normal network provides state-of-the-art predictions while the depth network, relying on much realistic smooth normal assumption, outperforms the traditional self-supervised depth prediction network by a large margin on the KITTI benchmark.
dc.description.statementofresponsibilityHuangying Zhan, Chamara Saroj Weerasekera, Ravi Garg, Ian Reid
dc.identifier.citationIEEE International Conference on Robotics and Automation, 2019 / Howard, A., Althoefer, K., Arai, F., Arrichiello, F., Caputo, B., Castellanos, J., Hauser, K., Isler, V., Kim, J., Liu, H., Oh, P., Santos, V., Scaramuzza, D., Ude, A., Voyles, R., Yamane, K., Okamura, A. (ed./s), vol.2019-May, pp.4811-4817
dc.identifier.doi10.1109/ICRA.2019.8793984
dc.identifier.isbn153866027X
dc.identifier.isbn9781538660263
dc.identifier.issn1050-4729
dc.identifier.issn2577-087X
dc.identifier.orcidGarg, R. [0000-0002-9422-8086]
dc.identifier.orcidReid, I. [0000-0001-7790-6423]
dc.identifier.urihttp://hdl.handle.net/2440/124480
dc.language.isoen
dc.publisherIEEE
dc.publisher.placePiscataway, NJ.
dc.relation.granthttp://purl.org/au-research/grants/arc/FL130100102
dc.relation.granthttp://purl.org/au-research/grants/arc/CE140100016
dc.relation.ispartofseriesIEEE International Conference on Robotics and Automation ICRA
dc.rights©2019 IEEE
dc.source.urihttps://ieeexplore.ieee.org/xpl/conhome/8780387/proceeding
dc.titleSelf-supervised learning for single view depth and surface normal estimation
dc.typeConference paper
pubs.publication-statusPublished

Files

Collections

Computer Science publications