search

wkom / VirtualPose

[ECCV 2022] Offical Pytorch implementation of "VirtualPose: Learning Generalizable 3D Human Pose Models from Virtual Data"
MIT License
66 stars 6 forks source link

Generalization #2

Open danielrogel opened 2 years ago

danielrogel commented 2 years ago

Does this method also generalize well to multi-view predictions ? , i.e generating heatmaps from several views and training in this manner to predict directly the 3d poses as in voxel pose (but the data is only synthetic only camera poses known I'm advance but no actual pairs available)

CHUNYUWANG commented 2 years ago

Yes, it is ready to be used in multi-view setting.

danielrogel commented 2 years ago

from the design of the code it seems like it is only tested on single view , am i right ?