MultiGO: Towards Multi-level Geometry Learning for Monocular 3D Textured Human Reconstruction

MultiGO is a model designed for monocular 3D human reconstruction, aiming to achieve high-quality reconstruction through multi-level geometry learning.

Before using MultiGO for inference, you need to download the model parameters. Please follow the steps below to obtain the required parameters:

1. Click the following link to download the model parameters:

   • Download Model Parameters

2. Extract the downloaded file and place the parameter files in the workspace directory of the project.

   • multigo/workspace/model.safetensors

1. To use MultiGO, you need to set up the appropriate PyTorch environment. Please refer to the following repositories to prepare your environment:

• LGM Repository
• SiTH Repository

Make sure to follow the instructions in these repositories to install the necessary dependencies and set up your environment correctly.

2. Download SMPL-X models and move them to the smpl_estimated_related/data/body_models folder. You should have the following data structure:

body_models
    └──smplx
        ├── SMPLX_NEUTRAL.pkl
        ├── SMPLX_NEUTRAL.npz
        ├── SMPLX_MALE.pkl
        ├── SMPLX_MALE.npz
        ├── SMPLX_FEMALE.pkl
        └── SMPLX_FEMALE.npz

Once your environment is set up, you can run the inference script by executing the following command:

bash infer.sh