动3D姿势和形状的幻象生成器以及基于2D和3D监督的编码器

Learning 3D Human Dynamics from VideoAngjoo Kanazawa∗, Jason Y. Zhang∗, Panna Felsen∗, Jitendra MalikUniversity of California, Berkeley{kanazawa,zhang.j,panna,malik}@eecs.berkeley.eduAbstractFrom an image of a person in action, we can easilyguess the 3D motion of the person in the immediate pastand future.This is because we have a mental model of3D human dynamics that we have acquired from observ-ing visual sequences of humans in motion.We presenta framework that can similarly learn a representation of3D dynamics of humans from video via a simple but ef-fective temporal encoding of image features. At test time,from video, the learned temporal representation give riseto smooth 3D mesh predictions. From a single image, ourmodel can recover the current 3D mesh as well as its 3Dpast and future motion.Our approach is designed so itcan learn from videos with 2D pose annotations in a semi-supervised manner. Though annotated data is always lim-ited, there are millions of videos uploaded daily on the In-ternet. In this work, we harvest this Internet-scale sourceof unlabeled data by training our model on unlabeled videowith pseudo-ground truth 2D pose obtained from an off-the-shelf 2D pose detector. Our experiments show that addingmore videos with pseudo-ground truth 2D pose monoton-ically improves 3D prediction performance. We evaluateour model on the recent challenging dataset of 3D Poses inthe Wild and obtain state-of-the-art performance on the 3Dprediction task without any fine-tuning. The project websitewith video can be found at https://akanazawa.github.io/human_dynamics/.1. IntroductionConsider the image of the baseball player mid-swing inFigure 1. Even though we only see a flat two-dimensionalpicture, we can infer the player’s 3D pose, as we can easilyimagine how his knees bend and arms extend in space. Fur-thermore, we can also infer his motion in the surroundingmoments as he swings the bat through. We can do this be-cause we have a mental model of 3D human dynamics thatwe have acquired from observing many examples of peoplein motion.∗ equal contributionInputPredictionsDifferentViewpointFigure 1: 3D motion prediction from a single image. We pro-pose a method that, given a single image of a person, predicts the3D mesh of the person’s body and also hallucinates the future andpast motion. Our method can learn from videos with only 2D poseannotations in a semi-supervised manner. Note our training setdoes not have any ground truth 3D pose sequences of batting mo-tion. Our model also produces smooth 3D predictions from videoinput.In this work, we present a computational framework thatcan similarly learn a model of 3D human dynamics fromvideo. Given a temporal sequence of images, we first ex-tract per-image features, and then train a simple 1D tem-poral encoder that learns a representation of 3D human dy-namics over a temporal context of image features. We forcethis representation to capture 3D human dynamics by pre-dicting not only the current 3D human pose and shape, butalso changes in pose in the nearby past and future frames.We transfer the learned 3D dynamics knowledge to staticimages by learning a hallucinator that can hallucinate thetemporal context representation from a single image fea-ture. The hallucinator is trained in a self-supervised mannerusing the actual output of the temporal encoder. Figure 2illustrates the overview of our training procedure.At test time, when the input is a video, the temporal en-15614f3DAAAB7nicbZDLSgMxFIbP1Fsdb1WXugiWgqsyowtdFnXhsgV7gXYomTTTxmaSIckIZeg7uBIUxK2P4Hu4cuejmF4W2vpD4OM/55Bz/jDhTBvP+3JyK6tr6xv5TXdre2d3r7B/0NAyVYTWieRStUKsKWeC1g0znLYSRXEcctoMh9eTevOBKs2kuDOjhAYx7gsWMYKNtRpRNzu/GXcLRa/sTYWWwZ9DsXL8UfsGgGq38NnpSZLGVBjCsdZt30tMkGFlGOF07JY6qaYJJkPcp22LAsdUB9l03TEqWaeHIqnsEwZNXffXRIZjrUdxaDtjbAZ6sTYx/6u1UxNdBhkTSWqoILOPopQjI9HkdtRjihLDRxYwUcwui8gAK0yMTci1KfiLNy9D46zsW67ZOK5gpjwcwQmcgg8XUIFbqEIdCNzDIzzDi5M4T86r8zZrzTnzmUP4I+f9B71ukVY=AAAB7nicbZDLSgMxFIbP1Fsdb1WXigRLwVWZqQtdFnXhsgV7gXYomTTTxmYyQ5IRytCle1eCgrj1EfoernwGX8L0stDWHwIf/zmHnPP7MWdKO86XlVlZXVvfyG7aW9s7u3u5/YO6ihJJaI1EPJJNHyvKmaA1zTSnzVhSHPqcNvzB9aTeeKBSsUjc6WFMvRD3BAsYwdpY9aCTnt+MOrm8U3SmQsvgziFfPh5Xvx9PxpVO7rPdjUgSUqEJx0q1XCfWXoqlZoTTkV1oJ4rGmAxwj7YMChxS5aXTdUeoYJwuCiJpntBo6tq/JlIcKjUMfdMZYt1Xi7WJ+V+tlejg0kuZiBNNBZl9FCQc6QhNbkddJinRfGgAE8nMsoj0scREm4Rsk4K7ePMy1EtF13DVxHEFM2XhCE7hDFy4gDLcQgVqQOAenuAFXq3YerberPdZa8aazxzCH1kfP50Zkrw=AAAB7nicbZDLSgMxFIbP1Fsdb1WXigRLwVWZqQtdFnXhsgV7gXYomTTTxmYyQ5IRytCle1eCgrj1EfoernwGX8L0stDWHwIf/zmHnPP7MWdKO86XlVlZXVvfyG7aW9s7u3u5/YO6ihJJaI1EPJJNHyvKmaA1zTSnzVhSHPqcNvzB9aTeeKBSsUjc6WFMvRD3BAsYwdpY9aCTnt+MOrm8U3SmQsvgziFfPh5Xvx9PxpVO7rPdjUgSUqEJx0q1XCfWXoqlZoTTkV1oJ4rGmAxwj7YMChxS5aXTdUeoYJwuCiJpntBo6tq/JlIcKjUMfdMZYt1Xi7WJ+V+tlejg0kuZiBNNBZl9FCQc6QhNbkddJinRfGgAE8nMsoj0scREm4Rsk4K7ePMy1EtF13DVxHEFM2XhCE7hDFy4gDLcQgVqQOAenuAFXq3YerberPdZa8aazxzCH1kfP50Zkrw=AAAB7nicbZBNS8NAEIYn9avWr6pHL4ul4KkketBjUQ8eK9gPaEPZbDft2s0m7E6EEvofPAkK4tX/48l/47bNQVtfWHh4Z4adeYNECoOu++0U1tY3NreK26Wd3b39g/LhUcvEqWa8yWIZ605ADZdC8SYKlLyTaE6jQPJ2ML6Z1dtPXBsRqwecJNyP6FCJUDCK1mqF/ezidtovV9yaOxdZBS+HCuRq9MtfvUHM0ogrZJIa0/XcBP2MahRM8mmp2ksNTygb0yHvWlQ04sbP5utOSdU6AxLG2j6FZO6Wfk1kNDJmEgW2M6I4Msu1mflfrZtieOVnQiUpcsUWH4WpJBiT2e1kIDRnKCcWKNPCLkvYiGrK0CZUsil4yzevQuu85lm+dyv16zyPIpzAKZyBB5dQhztoQBMYPMIzvMKbkzgvzrvzsWgtOPnMMfyR8/kDqy+PEQ==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AAAB6XicbZDLSgMxFIbPeK3jrerSTbAUXJUZN7oRi25ctmAv0A4lk562oZnMkGSEMvQJXAkK4rYP4wO4EN/G9LLQ1h8CH/9/DjnnhIng2njet7O2vrG5tZ3bcXf39g8O80fHdR2nimGNxSJWzZBqFFxizXAjsJkopFEosBEO76Z54xGV5rF8MKMEg4j2Je9xRo21qoNOvuCVvJnIKvgLKNx8uNfJ5MutdPKf7W7M0gilYYJq3fK9xAQZVYYzgWO32E41JpQNaR9bFiWNUAfZbNIxKVqnS3qxsk8aMnPdXx0ZjbQeRaGtjKgZ6OVsav6XtVLTuwoyLpPUoGTzj3qpICYm07VJlytkRowsUKa4HZawAVWUGXsc117BX955FeoXJd9y1SuUb2GuHJzCGZyDD5dQhnuoQA0YIDzBC7w6Q+fZeXPe56VrzqLnBP7ImfwAleyQVA==AAAB6XicbZBNS8NAEIYn9avGr6pHL4ul4KkkXvRY9OKxBfsBbSib7aRdutmE3Y1QQn+BJ0FBvPqTPPlv3LY5aOsLCw/vzLAzb5gKro3nfTulre2d3b3yvntweHR8Ujk96+gkUwzbLBGJ6oVUo+AS24Ybgb1UIY1Dgd1wer+od59QaZ7IRzNLMYjpWPKIM2qs1ZoMK1Wv7i1FNsEvoAqFmsPK12CUsCxGaZigWvd9LzVBTpXhTODcrQ0yjSllUzrGvkVJY9RBvtx0TmrWGZEoUfZJQ5au+2sip7HWszi0nTE1E71eW5j/1fqZiW6DnMs0MyjZ6qMoE8QkZHE2GXGFzIiZBcoUt8sSNqGKMmPDcW0K/vrNm9C5rvuWW161cVfkUYYLuIQr8OEGGvAATWgDA4RneIU3Z+q8OO/Ox6q15BQz5/BHzucPMeSNEw==resnetAAAB9XicbZBNS8NAEIYn9avGr6pHL8FS8FQSL3oRi148VrBVaINsthNdutnE3YlaQsF/4UlQEK/+DH+AJ/+N29aDXy8sPLwzw8y+USaFId//cEpT0zOzc+V5d2FxaXmlsrrWNmmuObZ4KlN9FjGDUihskSCJZ5lGlkQST6P+4ah+eo3aiFSd0CDDMGEXSsSCM7JW2CW8pUKjUUjD80rVr/tjeX8h+ILq/pu7dwcAzfPKe7eX8jxBRVwyYzqBn1FYME2CSxy6tW5uMGO8zy6wY1GxBE1YjK8eejXr9Lw41fYp8sau+22iYIkxgySynQmjS/O7NjL/q3VyinfDQqgsJ1R8sijOpUepN4rA6wmNnOTAAuNa2GM9fsk042SDcm0Kwe8//4X2dj2wfOxXGwcwURk2YBO2IIAdaMARNKEFHK7gHh7hyblxHpxn52XSWnK+Ztbhh5zXT311lJo=AAAB9XicbZDLSsNAFIYnXmu8VV26CZaCq5K40Y1YdOOygr1AG8pketIOnUzizIlaQp/DlaAobn0MH8CF+DZOLwtt/WHg4z/ncM78QSK4Rtf9thYWl5ZXVnNr9vrG5tZ2fme3puNUMaiyWMSqEVANgkuoIkcBjUQBjQIB9aB/MarXb0FpHstrHCTgR7QrecgZRWP5LYR7zBRoCThs5wtuyR3LmQdvCoWzD/s0efmyK+38Z6sTszQCiUxQrZuem6CfUYWcCRjaxVaqIaGsT7vQNChpBNrPxlcPnaJxOk4YK/MkOmPX/jWR0UjrQRSYzohiT8/WRuZ/tWaK4YmfcZmkCJJNFoWpcDB2RhE4Ha6AoRgYoExxc6zDelRRhiYo26Tgzf55HmpHJc/wlVson5OJcmSfHJBD4pFjUiaXpEKqhJEb8kCeyLN1Zz1ar9bbpHXBms7skT+y3n8AcHiWDg==AAAB9XicbZDLSsNAFIYnXmu8VV26CZaCq5K40Y1YdOOygr1AG8pketIOnUzizIlaQp/DlaAobn0MH8CF+DZOLwtt/WHg4z/ncM78QSK4Rtf9thYWl5ZXVnNr9vrG5tZ2fme3puNUMaiyWMSqEVANgkuoIkcBjUQBjQIB9aB/MarXb0FpHstrHCTgR7QrecgZRWP5LYR7zBRoCThs5wtuyR3LmQdvCoWzD/s0efmyK+38Z6sTszQCiUxQrZuem6CfUYWcCRjaxVaqIaGsT7vQNChpBNrPxlcPnaJxOk4YK/MkOmPX/jWR0UjrQRSYzohiT8/WRuZ/tWaK4YmfcZmkCJJNFoWpcDB2RhE4Ha6AoRgYoExxc6zDelRRhiYo26Tgzf55HmpHJc/wlVson5OJcmSfHJBD4pFjUiaXpEKqhJEb8kCeyLN1Zz1ar9bbpHXBms7skT+y3n8AcHiWDg==AAAB9XicbZBNS8NAEIY39avWr6pHL4ul4KkkXvRY9OKxgv2ANpTNdtIu3Wzi7kQtob/Dk6AgXv0xnvw3btsctPWFhYd3ZpjZN0ikMOi6305hbX1jc6u4XdrZ3ds/KB8etUycag5NHstYdwJmQAoFTRQooZNoYFEgoR2Mr2f19gNoI2J1h5ME/IgNlQgFZ2gtv4fwhJkGowCn/XLFrblz0VXwcqiQXI1++as3iHkagUIumTFdz03Qz5hGwSVMS9VeaiBhfMyG0LWoWATGz+ZXT2nVOgMaxto+hXTuln5NZCwyZhIFtjNiODLLtZn5X62bYnjpZ0IlKYLii0VhKinGdBYBHQgNHOXEAuNa2GMpHzHNONqgSjYFb/nPq9A6r3mWb91K/SrPo0hOyCk5Ix65IHVyQxqkSTi5J8/klbw5j86L8+58LFoLTj5zTP7I+fwBDHCSzQ==fmovieAAAB9nicbZA7SwNBEMfnfMb4ilpqcRgCVuHORsugjWUC5gFJDHubuWTJ3oPduZBw5HtYCQpia+n3sLLzo7h5FJr4h4Uf/5lhZv9eLIUmx/my1tY3Nre2MzvZ3b39g8Pc0XFNR4niWOWRjFTDYxqlCLFKgiQ2YoUs8CTWvcHttF4fotIiCu9pHGM7YL1Q+IIzMtaD32kRjigNoqHASSeXd4rOTPYquAvIl84+Kt8AUO7kPlvdiCcBhsQl07rpOjG1U6ZIcImTbKGVaIwZH7AeNg2GLEDdTmdnT+yCcbq2HynzQrJnbvbXRMoCrceBZzoDRn29XJua/9WaCfnX7VSEcUIY8vkiP5E2RfY0A7srFHKSYwOMK2GOtXmfKcbJJJU1KbjLf16F2mXRNVwxcdzAXBk4hXO4ABeuoAR3UIYqcFDwCM/wYo2sJ+vVepu3rlmLmRP4I+v9B8fTlXA=AAAB9nicbZDJSgNBEIZ74hbHLepRkcEQ8BRmvOgx6MVjAmaBZAw9nZqkSc9Cd01IGHL0HTwJCuLVY97Dk8/gS9hZDpr4Q8PHX1VU9e/Fgiu07S8js7a+sbmV3TZ3dvf2D3KHRzUVJZJBlUUikg2PKhA8hCpyFNCIJdDAE1D3+rfTen0AUvEovMdRDG5AuyH3OaOorQe/3UIYYhpEAw7jdi5vF+2ZrFVwFpAvnU4q349nk3I799nqRCwJIEQmqFJNx47RTalEzgSMzUIrURBT1qddaGoMaQDKTWdnj62CdjqWH0n9QrRmrvlrIqWBUqPA050BxZ5ark3N/2rNBP1rN+VhnCCEbL7IT4SFkTXNwOpwCQzFSANlkutjLdajkjLUSZk6BWf5z6tQuyw6mis6jhsyV5ackHNyQRxyRUrkjpRJlTAiyRN5Ia/G0Hg23oz3eWvGWMwckz8yPn4Ap36W1g==AAAB9nicbZDJSgNBEIZ74hbHLepRkcEQ8BRmvOgx6MVjAmaBZAw9nZqkSc9Cd01IGHL0HTwJCuLVY97Dk8/gS9hZDpr4Q8PHX1VU9e/Fgiu07S8js7a+sbmV3TZ3dvf2D3KHRzUVJZJBlUUikg2PKhA8hCpyFNCIJdDAE1D3+rfTen0AUvEovMdRDG5AuyH3OaOorQe/3UIYYhpEAw7jdi5vF+2ZrFVwFpAvnU4q349nk3I799nqRCwJIEQmqFJNx47RTalEzgSMzUIrURBT1qddaGoMaQDKTWdnj62CdjqWH0n9QrRmrvlrIqWBUqPA050BxZ5ark3N/2rNBP1rN+VhnCCEbL7IT4SFkTXNwOpwCQzFSANlkutjLdajkjLUSZk6BWf5z6tQuyw6mis6jhsyV5ackHNyQRxyRUrkjpRJlTAiyRN5Ia/G0Hg23oz3eWvGWMwckz8yPn4Ap36W1g==AAAB9nicbZBNS8NAEIY3ftb6VfXoZbEUPJXEix6LXjxWsB/QxrLZTtqlm03YnZSW0P/hSVAQr/4XT/4bt20O2vrCwsM7M8zsGyRSGHTdb2djc2t7Z7ewV9w/ODw6Lp2cNk2cag4NHstYtwNmQAoFDRQooZ1oYFEgoRWM7ub11hi0EbF6xGkCfsQGSoSCM7TWU9jrIkwwi+KxgFmvVHar7kJ0HbwcyiRXvVf66vZjnkagkEtmTMdzE/QzplFwCbNipZsaSBgfsQF0LCoWgfGzxdkzWrFOn4axtk8hXbjFXxMZi4yZRoHtjBgOzWptbv5X66QY3viZUEmKoPhyUZhKijGdZ0D7QgNHObXAuBb2WMqHTDOONqmiTcFb/fM6NK+qnuUHt1y7zfMokHNyQS6JR65JjdyTOmkQTjR5Jq/kzZk4L86787Fs3XDymTPyR87nD7WUkys=ΦtAAAB7nicbZC7SgNBFIbPxltcb1HBxmYwBKzCro2WQRvLBMwFkiXOTibJmNnZZeasEJa8g5WgILa+j5WNz+LkUmjiDwMf/zmHOecPEykMet6Xk1tb39jcym+7O7t7+weFw6OGiVPNeJ3FMtatkBouheJ1FCh5K9GcRqHkzXB0M603H7k2IlZ3OE54ENGBEn3BKFqr0akORRe7haJX9mYiq+AvoFg5qX3fA0C1W/js9GKWRlwhk9SYtu8lGGRUo2CST9xSJzU8oWxEB7xtUdGImyCbrTshJev0SD/W9ikkM9f9NZHRyJhxFNrOiOLQLNem5n+1dor9qyATKkmRKzb/qJ9KgjGZ3k56QnOGcmyBMi3ssoQNqaYMbUKuTcFfvnkVGhdl33LNxnENc+XhFM7gHHy4hArcQhXqwOABnuAFXp3EeXbenPd5a85ZzBzDHzkfP2YnkRI=AAAB7nicbZC7SgNBFIbPxltcb1HBxmYxBKzCro2WITaWCZgLJEuYncwmY2Znl5mzQljyDlaCgtja+Ba+gZWNz+LkUmjiDwMf/zmHOecPEsE1uu6XlVtb39jcym/bO7t7+weFw6OmjlNFWYPGIlbtgGgmuGQN5ChYO1GMRIFgrWB0Pa237pnSPJa3OE6YH5GB5CGnBI3V7NaGvIe9QtEtuzM5q+AtoFg5qX/z9+pHrVf47PZjmkZMIhVE647nJuhnRCGngk3sUjfVLCF0RAasY1CSiGk/m607cUrG6TthrMyT6Mxc+9dERiKtx1FgOiOCQ71cm5r/1Tophld+xmWSIpN0/lGYCgdjZ3q70+eKURRjA4QqbpZ16JAoQtEkZJsUvOWbV6F5UfYM100cVZgrD6dwBufgwSVU4AZq0AAKd/AAT/BsJdaj9WK9zltz1mLmGP7IevsBuSSSzg==AAAB7nicbZC7SgNBFIbPxltcb1HBxmYxBKzCro2WITaWCZgLJEuYncwmY2Znl5mzQljyDlaCgtja+Ba+gZWNz+LkUmjiDwMf/zmHOecPEsE1uu6XlVtb39jcym/bO7t7+weFw6OmjlNFWYPGIlbtgGgmuGQN5ChYO1GMRIFgrWB0Pa237pnSPJa3OE6YH5GB5CGnBI3V7NaGvIe9QtEtuzM5q+AtoFg5qX/z9+pHrVf47PZjmkZMIhVE647nJuhnRCGngk3sUjfVLCF0RAasY1CSiGk/m607cUrG6TthrMyT6Mxc+9dERiKtx1FgOiOCQ71cm5r/1Tophld+xmWSIpN0/lGYCgdjZ3q70+eKURRjA4QqbpZ16JAoQtEkZJsUvOWbV6F5UfYM100cVZgrD6dwBufgwSVU4AZq0AAKd/AAT/BsJdaj9WK9zltz1mLmGP7IevsBuSSSzg==AAAB7nicbZBNS8NAEIYn9avGr6pHL4ul4KkkXvRY9OKxgv2ANpTNdtOu3WzC7kQoof/Bk6AgXv0/nvw3btsctPWFhYd3ZtiZN0ylMOh5305pY3Nre6e86+7tHxweVY5P2ibJNOMtlshEd0NquBSKt1Cg5N1UcxqHknfCye283nni2ohEPeA05UFMR0pEglG0VrvfHIsBDipVr+4tRNbBL6AKhZqDyld/mLAs5gqZpMb0fC/FIKcaBZN85tb6meEpZRM64j2LisbcBPli3RmpWWdIokTbp5AsXPfXRE5jY6ZxaDtjimOzWpub/9V6GUbXQS5UmiFXbPlRlEmCCZnfToZCc4ZyaoEyLeyyhI2ppgxtQq5NwV+9eR3al3Xf8r1XbdwUeZThDM7hAny4ggbcQRNawOARnuEV3pzUeXHenY9la8kpZk7hj5zPH9Wbjy0=˜ΦtAAAB9nicbZDLSsNAFIZPvNZ4qwpu3ARLwVVJ3Oiy6MZlC/YCTayTybQdOpmEmROxhL6HK0FB3Pourtz4LE4vC239YeDjP+dwzvxhKrhG1/2yVlbX1jc2C1v29s7u3n7x4LCpk0xR1qCJSFQ7JJoJLlkDOQrWThUjcShYKxxeT+qtB6Y0T+QtjlIWxKQveY9Tgsa685GLiOV+bcDHXewWS27FncpZBm8Opepx/fseAGrd4qcfJTSLmUQqiNYdz00xyIlCTgUb22U/0ywldEj6rGNQkpjpIJ+ePXbKxomcXqLMk+hMXfvXRE5irUdxaDpjggO9WJuY/9U6GfYug5zLNEMm6WxRLxMOJs4kAyfiilEUIwOEKm6OdeiAKELRJGWbFLzFPy9D87ziGa6bOK5gpgKcwCmcgQcXUIUbqEEDKCh4ghd4tR6tZ+vNep+1rljzmSP4I+vjB9eGlMg=AAAB9nicbZDLSsNAFIYnXmu8VQU3bgZLwVVJ3Oiy1I3LFuwFmlgmk0k7dDIJMydiCX0PV4KCuBJ8CN/AlRufxelloa0/DHz85xzOmT9IBdfgOF/Wyura+sZmYcve3tnd2y8eHLZ0kinKmjQRieoERDPBJWsCB8E6qWIkDgRrB8OrSb19x5TmibyBUcr8mPQljzglYKxbD7gIWe7VB3zcg16x5FScqfAyuHMoVY8b3/yt9lHvFT+9MKFZzCRQQbTuuk4Kfk4UcCrY2C57mWYpoUPSZ12DksRM+/n07DEuGyfEUaLMk4Cnrv1rIiex1qM4MJ0xgYFerE3M/2rdDKJLP+cyzYBJOlsUZQJDgicZ4JArRkGMDBCquDkW0wFRhIJJyjYpuIt/XobWecU13DBx1NBMBXSCTtEZctEFqqJrVEdNRJFCD+gJPVv31qP1Yr3OWles+cwR+iPr/QcqkpaE

下载后可阅读完整内容，剩余1页未读，立即下载