ShapeBoost: Boosting Human Shape Estimation with Part-based Parameterization and Clothing-preserving Augmentation

The code is adapted from Original Hybrik.

Examples

Installation instructions

# 1. Create a conda virtual environment.
conda create -n shapeboost python=3.10 -y
conda activate shapeboost

# 2. Install PyTorch, you may switch to other pytorch and cuda versions if necessary
conda install -c "nvidia/label/cuda-11.8.0" cuda-toolkit
pip install torch==2.4.0 torchvision torchaudio --index-url https://download.pytorch.org/whl/cu118

# Install Pytorch3D
conda install -c iopath iopath
conda install -c bottler nvidiacub
conda install pytorch3d -c pytorch3d

# 4. Install
python setup.py develop

Pretrained models

The pretrained model is available on Models.

The extra files are available at here and here.

Code Structure

ShapeBoost/
├── shapeboost/               
├── examples/                   
├── extra_files/                 
├── model_files/
│   └── smpl_v1.1.0
│   └── ...
└── ...

Demo

./scripts/demo.sh configs/hrw48_cam_2x_sratio_semi_analytical.yaml data/model_14_analytical_finetune.pth

Train from scratch

./scripts/train_smpl_shape_ddp.sh train_shapeboost configs/hrw48_cam_2x_sratio_semi_analytical.yaml

Evaluation

./scripts/validate_smpl_shape_ddp.sh --cfg configs/NIKI_twistswing.yaml --ckpt {CKPT}