search

yanx27 / EverybodyDanceNow_reproduce_pytorch

Everybody dance now reproduced in pytorch
MIT License
607 stars 173 forks source link

questions about training #61

Open zrhyst23 opened 5 years ago

zrhyst23 commented 5 years ago

Excuse me ,i think there's something strange here.Traning with only one video,Does each video have to train the model once?I

zhanjichao commented 4 years ago

I have the same problem. Do you figure out it now?

jbartolozzi commented 4 years ago

Training will just run pix2pixHD on the "target" skeleton and image files. Then the inference will use the "source" skeleton to produce images of the target. You only train on the "target" skeleton pose and images.

zhanjichao commented 4 years ago

Thanks! I got it. :)

获取 Outlook for Androidhttps://aka.ms/ghei36

From: James Bartolozzi notifications@github.com Sent: Friday, January 10, 2020 4:56:23 AM To: CUHKSZ-TQL/EverybodyDanceNow_reproduce_pytorch EverybodyDanceNow_reproduce_pytorch@noreply.github.com Cc: jichao-zhan zhanjich@outlook.com; Comment comment@noreply.github.com Subject: Re: [CUHKSZ-TQL/EverybodyDanceNow_reproduce_pytorch] questions about training (#61)

Training will just run pix2pixHD on the "target" skeleton and image files. Then the inference will use the "source" skeleton to produce images of the target. You only train on the "target" skeleton pose and images.

― You are receiving this because you commented. Reply to this email directly, view it on GitHubhttps://github.com/CUHKSZ-TQL/EverybodyDanceNow_reproduce_pytorch/issues/61?email_source=notifications&email_token=ALHOCUXUDQXU7XYQBYC7KJ3Q46FPPA5CNFSM4IVCAET2YY3PNVWWK3TUL52HS4DFVREXG43VMVBW63LNMVXHJKTDN5WW2ZLOORPWSZGOEIRX6BY#issuecomment-572751623, or unsubscribehttps://github.com/notifications/unsubscribe-auth/ALHOCUUG7HTTXHMHI774LHDQ46FPPANCNFSM4IVCAETQ.

lauramercier commented 3 years ago

Is it meaning that I can not use the model trained to generate a video with a target video and a source pose(the npy file)?