# SATO: Stable Text-to-Motion Framework [Wenshuo chen*](https://github.com/shurdy123), [Hongru Xiao*](https://github.com/Hongru0306), [Erhang Zhang*](https://github.com/zhangerhang), [Lijie Hu](https://sites.google.com/view/lijiehu/homepage), [Lei Wang](https://leiwangr.github.io/), [Mengyuan Liu](), [Chen Chen](https://www.crcv.ucf.edu/chenchen/) [![Website shields.io](https://img.shields.io/website?url=http%3A//poco.is.tue.mpg.de)](https://sato-team.github.io/Stable-Text-to-Motion-Framework/) [![YouTube Badge](https://img.shields.io/badge/YouTube-Watch-red?style=flat-square&logo=youtube)]() [![arXiv](https://img.shields.io/badge/arXiv-2308.12965-00ff00.svg)]() ## Existing Challenges A fundamental challenge inherent in text-to-motion tasks stems from the variability of textual inputs. Even when conveying similar or the same meanings and intentions, texts can exhibit considerable variations in vocabulary and structure due to individual user preferences or linguistic nuances. Despite the considerable advancements made in these models, we find a notable weakness: all of them demonstrate instability in prediction when encountering minor textual perturbations, such as synonym substitutions. In the following demonstration, we showcase the instability of predictions generated by the previous method when presented with different user inputs conveying identical semantic meaning.

Original text: A man kicks something or someone with his left leg.
T2M-GPT MDM MoMask
gif gif gif
Perturbed text: A human boots something or someone with his left leg.
T2M-GPT MDM MoMask
gif gif gif

## Motivation ![motivation](images/motivation.png) The model's inconsistent outputs are accompanied by unstable attention patterns. We further elucidate the aforementioned experimental findings: When perturbed text is inputted, the model exhibits unstable attention, often neglecting critical text elements necessary for accurate motion prediction. This instability further complicates the encoding of text into consistent embeddings, leading to a cascade of consecutive temporal motion generation errors. ## Visualization

Original text: person is walking normally in a circle.
T2M-GPT MDM MoMask SATO
gif gif gif gif
Perturbed text: human is walking usually in a loop.
T2M-GPT MDM MoMask SATO
gif gif gif gif

Explanation: T2M-GPT, MDM, and MoMask all don't walk in a loop.

Original text: a person uses his right arm to help himself to stand up.
T2M-GPT MDM MoMask SATO
gif gif gif gif
Perturbed text: A human utilizes his right arm to help himself to stand up.
T2M-GPT MDM MoMask SATO
gif gif gif gif

Explanation: T2M-GPT, MDM, and MoMask all lack the action of transitioning from squatting to standing up, resulting in a catastrophic error.

## How to Use the Code * [1. Setup and Installation](#setup) * [2.Dependencies](#Dependencies) * [3. Quick Start](#quickstart) * [4. Datasets](#datasets) * [4. Train](#train) * [5. Evaluation](#eval) * [6. Acknowledgments](#acknowledgements) ## Setup and Installation Clone the repository: ```shell git clone https://github.com/sato-team/Stable-Text-to-motion-Framework.git ``` Create fresh conda environment and install all the dependencies: ``` conda env create -f environment.yml conda activate SATO ``` The code was tested on Python 3.8 and PyTorch 1.8.1. ## Dependencies ```shell bash dataset/prepare/download_extractor.sh bash dataset/prepare/download_glove.sh ``` ## **Quick Start** A quick reference guide for using our code is provided in quickstart.ipynb. ## Datasets We are using two 3D human motion-language dataset: HumanML3D and KIT-ML. For both datasets, you could find the details as well as download [link](https://github.com/EricGuo5513/HumanML3D). We perturbed the input texts based on the two datasets mentioned. You can access the perturbed text dataset through the following [link](https://drive.google.com/file/d/1XLvu2jfG1YKyujdANhYHV_NfFTyOJPvP/view?usp=sharing). Take HumanML3D for an example, the dataset structure should look like this: ``` ./dataset/HumanML3D/ ├── new_joint_vecs/ ├── texts/ # You need to replace the 'texts' folder in the original dataset with the 'texts' folder from our dataset. ├── Mean.npy ├── Std.npy ├── train.txt ├── val.txt ├── test.txt ├── train_val.txt └── all.txt ``` ### **Train** We will release the training code soon. ### **Evaluation** You can download the pretrained models in this [link](https://drive.google.com/drive/folders/1rs8QPJ3UPzLW4H3vWAAX9hJn4ln7m_ts?usp=sharing). ```shell python eval_t2m.py --resume-pth pretrained/net_best_fid.pth --clip_path pretrained/clip_best_fid.pth ``` ## Acknowledgements We appreciate helps from : - Open Source Code：[T2M-GPT](https://github.com/Mael-zys/T2M-GPT), [MoMask ](https://github.com/EricGuo5513/momask-codes), [MDM](https://guytevet.github.io/mdm-page/), etc. - [Hongru Xiao](https://github.com/Hongru0306), [Erhang Zhang](https://github.com/zhangerhang), [Lijie Hu](https://sites.google.com/view/lijiehu/homepage), [Lei Wang](https://leiwangr.github.io/), [Mengyuan Liu](), [Chen Chen](https://www.crcv.ucf.edu/chenchen/) for discussions and guidance throughout the project, which has been instrumental to our work. - [Zhen Zhao](https://github.com/Zanebla) for project website. - If you find our work helpful, we would appreciate it if you could give our project a star! ## Citing If you find this code useful for your research, please consider citing the following paper: ```bibtex ```