TapMo: Shape-aware Motion Generation of Skeleton-free Characters

This is the code for TapMo: Shape-aware Motion Generation of Skeleton-free Characters by Jiaxu Zhang, et al.

TapMo is a text-based animation pipeline for generating motion in a wide variety of skeleton-free characters.

Inference code
Training code

Prerequisites

Python >= 3.7
Pytorch >= 1.4
Pytorch-geometric

Quick Start

1. Conda environment

conda create python=3.8 --name tapmo
conda activate tapmo

2. Install dependencies

Install the packages in requirements.txt and install PyTorch 2.1.0

pip install -r requirements.txt

3. Download the datasets and the requriements

Download the processed datasets and the requriements from Google dirve

cd TapMo
unzip datasets.zip -d ./
unzip weights.zip -d ./
unzip deps.zip -d ./shape_diffusion

4. Run

cd shape_diffusion
python3 -m sample.generate_handle_motion --model_path ../weights/diffusion_model_latest.pt --arch trans_dec --emb_trans_dec False --dataset t6d_mixrig --char_feature_path ../demo/shape_features/001.npy --save_path ../demo/motion/motion_ --text_prompt "walk forward and turn right."

cd handle_predictor
python -m motion_to_mesh --ckpt_path ../weights/handle_predictor_latest.pth --motion_path ../demo/motion/motion_0.npz --tgt_mesh_path ../demo/mesh/001.obj --save_dir ../demo/results/001

Citation

Please cite our paper if you use this repository:

@inproceedings{zhang2024tapmo,
    title = {TapMo: Shape-aware Motion Generation of Skeleton-free Characters},
    author = {Zhang, Jiaxu and Huang, Shaoli and Tu, Zhigang and Chen, Xin and Zhan, Xiaohang and Yu, Gang and Shan, Ying},
    booktitle = {The Twelfth International Conference on Learning Representations ({ICLR})},
    year = {2024},
}

Credit

We borrowed part of the codes from the following projects:

https://github.com/zycliao/skeleton-free-pose-transfer

https://github.com/GuyTevet/motion-diffusion-model

Kebii/TapMo