ImageNet-trained CNNs are biased towards texture; increasing shape bias improves accuracy and robustness

dc.contributor.author	Bethge, Matthias
dc.contributor.author	Wichmann, Felix A.
dc.contributor.author	Geirhos, Robert
dc.contributor.author	Rubisch, Patricia
dc.contributor.author	Michaelis, Claudio
dc.contributor.author	Brendel, Wieland
dc.date.accessioned	2020-11-16T15:49:38Z
dc.date.available	2020-11-16T15:49:38Z
dc.date.issued	2018-11-29
dc.identifier.uri	http://hdl.handle.net/10900/109613
dc.language.iso	en	de_DE
dc.relation.uri	https://arxiv.org/abs/1811.12231	de_DE
dc.subject.ddc	004	de_DE
dc.title	ImageNet-trained CNNs are biased towards texture; increasing shape bias improves accuracy and robustness	de_DE
dc.type	Preprint	de_DE
utue.personen.roh	Geirhos, Robert
utue.personen.roh	Rubisch, Patricia
utue.personen.roh	Michaelis, Claudio
utue.personen.roh	Bethge, Matthias
utue.personen.roh	Wichmann, Felix A.
utue.personen.roh	Brendel, Wieland
dcterms.isPartOf.ZSTitelID	arXiv	de_DE
dcterms.isPartOf.ZS-Volume	1811.12231	de_DE