Datasets:
ArXiv:
License:
Update README.md
Browse files
README.md
CHANGED
|
@@ -1,3 +1,62 @@
|
|
| 1 |
-
---
|
| 2 |
-
license: apache-2.0
|
| 3 |
-
---
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
license: apache-2.0
|
| 3 |
+
---
|
| 4 |
+
## π Dataset Usage
|
| 5 |
+
To facilitate researchers, we provide the processed streaming 272-dim Motion Representation of [BABEL](https://babel.is.tue.mpg.de/) dataset in this Hugging Face repo.
|
| 6 |
+
|
| 7 |
+
NOTE: We process the original BABEL dataset to support training of streaming motion generation.
|
| 8 |
+
e.g. If there is a motion sequence A, annotated as (A1, A2, A3, A4) in BABEL dataset, each subsequence has text description: (A1_t, A2_t, A3_t, A4_t).
|
| 9 |
+
|
| 10 |
+
Then, our BABEL-stream is constructed as:
|
| 11 |
+
|
| 12 |
+
seq1: (A1, A2) --- seq1_text: (A1_t*A2_t#A1_length)
|
| 13 |
+
|
| 14 |
+
seq2: (A2, A3) --- seq2_text: (A2_t*A3_t#A2_length)
|
| 15 |
+
|
| 16 |
+
seq3: (A3, A4) --- seq3_text: (A3_t*A4_t#A3_length)
|
| 17 |
+
|
| 18 |
+
Here, * and # is separation symbol, A1_length means the number of frames of subsequence A1.
|
| 19 |
+
|
| 20 |
+
Motions are resampled into 30 FPS.
|
| 21 |
+
|
| 22 |
+
The dataset is organized as:
|
| 23 |
+
```
|
| 24 |
+
./
|
| 25 |
+
βββ train_stream
|
| 26 |
+
βββ seq1.npy
|
| 27 |
+
...
|
| 28 |
+
βββ train_stream_text
|
| 29 |
+
βββ seq1.txt
|
| 30 |
+
...
|
| 31 |
+
βββ val_stream
|
| 32 |
+
βββ seq1.npy
|
| 33 |
+
...
|
| 34 |
+
βββ val_stream_text
|
| 35 |
+
βββ seq1.txt
|
| 36 |
+
...
|
| 37 |
+
```
|
| 38 |
+
|
| 39 |
+
βοΈβοΈβοΈ The processed data is solely for academic purposes. Make sure you read through the [BABEL License](https://babel.is.tue.mpg.de/license.html).
|
| 40 |
+
|
| 41 |
+
## π Paper & Project Page & Code
|
| 42 |
+
* [Arxiv Paper](https://arxiv.org/abs/2503.15451)
|
| 43 |
+
* [Project Page](https://zju3dv.github.io/MotionStreamer/)
|
| 44 |
+
* [Code](https://github.com/zju3dv/MotionStreamer)
|
| 45 |
+
|
| 46 |
+
## π Processing script
|
| 47 |
+
For more details of how to obtain the 272-dim motion representation, as well as other useful tools (e.g., Visualization and Conversion to BVH format), please refer to our [GitHub repo](https://github.com/Li-xingXiao/272-dim-Motion-Representation).
|
| 48 |
+
|
| 49 |
+
## πΉ Acknowledgement
|
| 50 |
+
This repository builds upon the following awesome datasets and projects:
|
| 51 |
+
- [BABEL](https://babel.is.tue.mpg.de/)
|
| 52 |
+
|
| 53 |
+
## π€πΌ Citation
|
| 54 |
+
If our project is helpful for your research, please consider citing :
|
| 55 |
+
```
|
| 56 |
+
@article{xiao2025motionstreamer,
|
| 57 |
+
title={MotionStreamer: Streaming Motion Generation via Diffusion-based Autoregressive Model in Causal Latent Space},
|
| 58 |
+
author={Xiao, Lixing and Lu, Shunlin and Pi, Huaijin and Fan, Ke and Pan, Liang and Zhou, Yueer and Feng, Ziyong and Zhou, Xiaowei and Peng, Sida and Wang, Jingbo},
|
| 59 |
+
journal={arXiv preprint arXiv:2503.15451},
|
| 60 |
+
year={2025}
|
| 61 |
+
}
|
| 62 |
+
```
|