Visual Question Answering as a meta learning task

Simple item page
dc.contributor.authorTeney, D.
dc.contributor.authorVan Den Hengel, A.
dc.contributor.conference15th European Conference on Computer Vision (ECCV 2018) (8 Sep 2018 - 14 Sep 2018 : Munich)
dc.contributor.editorFerrari, V.
dc.contributor.editorHebert, M.
dc.contributor.editorSminchisescu, C.
dc.contributor.editorWeiss, Y.
dc.date.issued2018
dc.description.abstractThe predominant approach to Visual Question Answering (VQA) demands that the model represents within its weights all of the information required to answer any question about any image. Learning this information from any real training set seems unlikely, and representing it in a reasonable number of weights doubly so. We propose instead to approach VQA as a meta learning task, thus separating the question answering method from the information required. At test time, the method is provided with a support set of example questions/answers, over which it reasons to resolve the given question. The support set is not fixed and can be extended without retraining, thereby expanding the capabilities of the model. To exploit this dynamically provided information, we adapt a state-of-the-art VQA model with two techniques from the recent meta learning literature, namely prototypical networks and meta networks. Experiments demonstrate the capability of the system to learn to produce completely novel answers (i.e. never seen during training) from examples provided at test time. In comparison to the existing state of the art, the proposed method produces qualitatively distinct results with higher recall of rare answers, and a better sample efficiency that allows training with little initial data. More importantly, it represents an important step towards vision-and-language methods that can learn and reason on-the-fly.
dc.description.statementofresponsibilityDamien Teney and Anton van den Hengel
dc.identifier.citationLecture Notes in Artificial Intelligence, 2018 / Ferrari, V., Hebert, M., Sminchisescu, C., Weiss, Y. (ed./s), vol.11219 LNCS, pp.229-245
dc.identifier.doi10.1007/978-3-030-01267-0_14
dc.identifier.isbn9783030012663
dc.identifier.issn0302-9743
dc.identifier.issn1611-3349
dc.identifier.orcidTeney, D. [0000-0003-2130-6650]
dc.identifier.orcidVan Den Hengel, A. [0000-0003-3027-8364]
dc.identifier.urihttp://hdl.handle.net/2440/116282
dc.language.isoen
dc.publisherSpringer
dc.relation.ispartofseriesLecture Notes in Computer Science; 11219
dc.rights© Springer Nature Switzerland AG 2018
dc.source.urihttps://doi.org/10.1007/978-3-030-01267-0_14
dc.titleVisual Question Answering as a meta learning task
dc.typeConference paper
pubs.publication-statusPublished

Files

Collections

Australian Institute for Machine Learning publications