Mueller, FranziskaDavis, MicahBernard, FlorianSotnychenko, OleksandrVerschoor, MickealOtaduy, Miguel A.Casas, DanTheobalt, Christian2020-04-202020-04-202019ACM Transactions on Graphics July 2019 Article No.: 49 https://doi.org/10.1145/3306346.33229581557-7368http://hdl.handle.net/10115/16789Wepresentanovelmethodforreal-timeposeandshapereconstructionof twostronglyinteractinghands.Ourapproachisthefirsttwo-handtracking solutionthatcombinesanextensivelistoffavorableproperties,namelyitis marker-less,usesasingleconsumer-leveldepthcamera,runsinrealtime, handlesinter-andintra-handcollisions,andautomaticallyadjuststothe user’shandshape.Inordertoachievethis,weembedarecentparametric handposeandshapemodelandadensecorrespondencepredictorbasedon adeepneuralnetworkintoasuitableenergyminimizationframework.For trainingthecorrespondencepredictionnetwork,wesynthesizeatwo-hand dataset based on physical simulations that includes both hand pose and shapeannotationswhileatthesametimeavoidinginter-handpenetrations. Toachievereal-timerates,wephrasethemodelfittingintermsofanonlinear least-squaresproblemsothattheenergycanbeoptimizedbasedonahighly efficient GPU-based Gauss-Newton optimizer. We show state-of-the-art resultsinscenesthatexceedthecomplexityleveldemonstratedbypreviousengAtribución-NoComercial-CompartirIgual 4.0 Internacionalhttp://creativecommons.org/licenses/by-nc-sa/4.0/Computing methodologiesArtificial intelligenceMachine learningInformáticainfo:eu-repo/semantics/preprint10.1145/3306346.3322958info:eu-repo/semantics/openAccess1203.04 Inteligencia Artificial