To deploy and run FastAnimate, run the following scripts:
conda create -n fastanimate
pip install -r requirements.txt
conda activate fastanimate
Then, download and compile diff-gaussian-rasterization, kaolin and nvidiffrast.
- SMPL/SMPL-X model: register and download SMPL and SMPL-X, and put these files in
/smpl_related/to replace the pseudo files. The folder should have the following structure:
smpl_related
└── models
└──smpl
├── ...
├── ...
└── ...
└──smplx
├── ...
├── ...
└── ...
└── smpl_data
├── ...
├── ...
└── ...
- Data: register and download X-Humans
CUDA_VISIBLE_DEVICES=0,1,2,3 nohup accelerate launch --config_file acc_configs/gpu4.yaml train.py big --workspace workspace_train --resume ./path_to_your_stage_1_models
CUDA_VISIBLE_DEVICES=0 python test.py big --resume ./path_to_your_stage_2_models --workspace workspace_test
This project is built on source codes shared by LGM, SiTH, MVDream and MultiGo.