search

lizhe00 / AvatarCap

Code of [ECCV 2022] "AvatarCap: Animatable Avatar Conditioned Monocular Human Volumetric Capture"
http://www.liuyebin.com/avatarcap/avatarcap.html
179 stars 22 forks source link

About preprocessing training data #23

Closed green-slime closed 1 year ago

green-slime commented 1 year ago

Hi! I got a little confused when preprocessing training data——EasyMocap needs videos of multiple views as input, so if i understand it correctly, I should use multiple DSLR to record videos and use EasyMocap to get SMPL parameters of each frame, is it right? But I don't know how to get 3d scan files this way…… Sorry I'm a newer in 3DV, thank you and looking forward to your help. :sob: Hi! If I understand it correctly, firstly we take photos by a group of DSLR, then on one hand we use these photos to get multi-view 3D reconstruction result, on the other hand we use them as 1-frame videos as the input of EasyMocap. I wonder if I get it or not...

lizhe00 commented 1 year ago

Sorry for the late reply. If you want to fit SMPL to a 3D scan, you can render the 3D scan to multiview images, then feed these images to EasyMocap to get SMPL parameters.

green-slime commented 1 year ago

Sorry for the late reply. If you want to fit SMPL to a 3D scan, you can render the 3D scan to multiview images, then feed these images to EasyMocap to get SMPL parameters.

Thanks a lot for your help! But EasyMocap requires camera parameters as input, so can we still get correct camera parameters when rendering 3D scan to multiview images?

lizhe00 commented 1 year ago

Yes, you shall input the rendered multiview images and corresponding camera parameters to EasyMocap.

green-slime commented 1 year ago

Thanks so much! I'll try it!