search

choyingw / SynergyNet

3DV 2021: Synergy between 3DMM and 3D Landmarks for Accurate 3D Facial Geometry
MIT License
377 stars 57 forks source link

Label of train_aug_120x120 #11

Closed vodanhbk95 closed 2 years ago

vodanhbk95 commented 2 years ago

When I load file './3dmm_data/param_all_norm_v201.pkl' , I found that it has shape ([636252, 102]). I guess that 636252 is the number of image for training but I still dont understand that what 102 is? Can you explain to me about how you can create your training set?

choyingw commented 2 years ago

Hi, the number of 636252 is the training data size. The 102 in order contains 12 pose, 40 shape, 10 expression, 40 textures(not used in main project, only for comparison with textures synthesized from 3DMM). So you can use the first 62 parameters to reconstruct the 3d face.

vodanhbk95 commented 2 years ago

Big thanks. But I still wondered how you create the first 62 parameters. Do you have the code that did it?

choyingw commented 2 years ago

This follows 3DDFA/ 3DDFA-V2 (https://github.com/cleardusk/3DDFA) whitening: subtract the original 3DMM parameter mean and divide it by the standard deviation to get the 62 parameters.