Handle URI:
http://hdl.handle.net/10754/626720
Title:
Pixels to Graphs by Associative Embedding
Authors:
Newell, Alejandro; Deng, Jia
Abstract:
Graphs are a useful abstraction of image content. Not only can graphs represent details about individual objects in a scene but they can capture the interactions between pairs of objects. We present a method for training a convolutional neural network such that it takes in an input image and produces a full graph. This is done end-to-end in a single stage with the use of associative embeddings. The network learns to simultaneously identify all of the elements that make up a graph and piece them together. We benchmark on the Visual Genome dataset, and report a Recall@50 of 9.7% compared to the prior state-of-the-art at 3.4%, a nearly threefold improvement on the challenging task of scene graph generation.
Publisher:
arXiv
KAUST Grant Number:
OSR-2015-CRG4-2639
Issue Date:
22-Jun-2017
ARXIV:
arXiv:1706.07365
Type:
Preprint
Additional Links:
http://arxiv.org/abs/1706.07365v1; http://arxiv.org/pdf/1706.07365v1
Appears in Collections:
Publications Acknowledging KAUST Support

Full metadata record

DC FieldValue Language
dc.contributor.authorNewell, Alejandroen
dc.contributor.authorDeng, Jiaen
dc.date.accessioned2018-01-04T07:51:42Z-
dc.date.available2018-01-04T07:51:42Z-
dc.date.issued2017-06-22en
dc.identifier.urihttp://hdl.handle.net/10754/626720-
dc.description.abstractGraphs are a useful abstraction of image content. Not only can graphs represent details about individual objects in a scene but they can capture the interactions between pairs of objects. We present a method for training a convolutional neural network such that it takes in an input image and produces a full graph. This is done end-to-end in a single stage with the use of associative embeddings. The network learns to simultaneously identify all of the elements that make up a graph and piece them together. We benchmark on the Visual Genome dataset, and report a Recall@50 of 9.7% compared to the prior state-of-the-art at 3.4%, a nearly threefold improvement on the challenging task of scene graph generation.en
dc.publisherarXiven
dc.relation.urlhttp://arxiv.org/abs/1706.07365v1en
dc.relation.urlhttp://arxiv.org/pdf/1706.07365v1en
dc.titlePixels to Graphs by Associative Embeddingen
dc.typePreprinten
dc.contributor.institutionComputer Science and Engineering, University of Michigan, Ann Arboren
dc.identifier.arxividarXiv:1706.07365en
kaust.grant.numberOSR-2015-CRG4-2639en
All Items in KAUST are protected by copyright, with all rights reserved, unless otherwise indicated.