search

gulvarol / bodynet

BodyNet: Volumetric Inference of 3D Human Body Shapes, ECCV 2018
http://www.di.ens.fr/willow/research/bodynet
MIT License
261 stars 42 forks source link

camera Intrinsic parameters #3

Closed rafikg closed 5 years ago

rafikg commented 5 years ago

@gulvarol Thanks for the great work. I read that you have supposed that the camera intrinsic parameters are known during the training. So, what did you do during the inference mode? Do you need camera intrinsic parameters during inference?

gulvarol commented 5 years ago

For shape: I had an approximate projection, so I just scale the volume with a constant that I compute from the 2D segmentation map: https://github.com/gulvarol/bodynet/blob/master/fitting/fit_surreal.py#L129 This is undoing the operation that was done to the ground truth volume at training: https://github.com/gulvarol/bodynet/blob/master/training/util/voxelutils.lua#L3