Text-to-motion synthesis is a crucial task in computer vision. Existing methods are limited in their universality, as they are tailored for single-person or two-person scenarios and can not be applied to generate motions for more individuals. To achieve the number-free motion synthesis, this paper reconsiders motion generation and proposes to unify the single and multi-person motion by the conditional motion distribution. Furthermore, a generation module and an interaction module are designed for our FreeMotion framework to decouple the process of conditional motion generation and finally support the number-free motion synthesis. Besides, based on our framework, the current single-person motion spatial control method could be seamlessly integrated, achieving precise control of multi-person motion. Extensive experiments demonstrate the superior performance of our method and our capability to infer single and multi-human motions simultaneously.
- [✓] Release the FreeMotion training.
- [✓] Release the FreeMotion evaluation.
- [] Release the separate_annots dataset.
- [] Release the FreeMotion checkpoints.
conda create python=3.8 --name freemotion
conda activate freemotion
pip install -r requirements.txt
bash prepare/download_evaluation_model.sh
Download the data from InterGen Webpage. And put them into ./data/.
Download the data from Ours Webpage And put it into ./data/.
<DATA-DIR>
./annots //Natural language annotations where each file consisting of three sentences.
./motions //Raw motion data standardized as SMPL which is similiar to AMASS.
./motions_processed //Processed motion data with joint positions and rotations (6D representation) of SMPL 22 joints kinematic structure.
./split //Train-val-test split.
./separate_annots //Annotations for each person's motion
sh train_single.sh
sh train_inter.sh
sh test.sh
If you find our code or paper helps, please consider citing:
@article{fan2024freemotion,
title={FreeMotion: A Unified Framework for Number-free Text-to-Motion Synthesis},
author={Ke Fan and Junshu Tang and Weijian Cao and Ran Yi and Moran Li and Jingyu Gong and Jiangning Zhang and Yabiao Wang and Chengjie Wang and Lizhuang Ma},
year={2024},
eprint={2405.15763},
archivePrefix={arXiv},
primaryClass={cs.CV}
}
Thanks to interhuman,MotionGPT, our code is partially borrowing from them.
This work is licensed under a Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License.