# The following items are already included:
mean_deform.pt # for Hair/shoulder deformation model
u_full.pt # Hair/shoulder deformation model
transform.pkl # downsample/upsample mesh from 5024 vertices to 314 vertices, for Spiralnet
head_temmplate.obj # FLAME face model
# Yo need to download the following items
cvthead.pth # our pretrained CVT Head on VoxCeleb1
79999_iter.pth # Face Parsing model from https://github.com/VisionSystemsInc/face-parsing.PyTorch
linear_hair.pth # Hair/shoulder deformation model from ROME
rome.pth # pre-trained ROME, we will use its CNN feature extractor
generic_model.pkl # FLAME model, obtained from https://flame.is.tue.mpg.de/
resnet50_scratch_weight.pth # pretrained ResNet-50 on VGGFace2, for face identity loss
deca_model.tar # pre-trained DECA
You can download all files by:
bash fetch_data.sh
This scripts will automatically download all required files
In case it doesn't work, you can also download each file separately from their project pages:
You can also find this file at: face-parsing.PyTorch
-
Linear deformation model linear_hair.pth
-
ROME rome.pth
You can also find all these files at: rome
- resnet50_scratch_weight.pth You can also find it at: VGGFace2-pytorch