Add dataset card and link to paper
#2
by nielsr HF Staff - opened
README.md
CHANGED
|
@@ -1,3 +1,78 @@
|
|
| 1 |
-
---
|
| 2 |
-
license: mit
|
| 3 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
license: mit
|
| 3 |
+
task_categories:
|
| 4 |
+
- image-to-video
|
| 5 |
+
---
|
| 6 |
+
|
| 7 |
+
# FlexiMMT Dataset
|
| 8 |
+
|
| 9 |
+
This repository contains the benchmark data for **FlexiMMT**, the first implicit image-to-video (I2V) motion transfer framework that explicitly enables multi-object, multi-motion transfer.
|
| 10 |
+
|
| 11 |
+
The dataset was presented in the paper [Let Your Image Move with Your Motion! -- Implicit Multi-Object Multi-Motion Transfer](https://huggingface.co/papers/2603.01000).
|
| 12 |
+
|
| 13 |
+
[**Project Page**](https://ethan-li123.github.io/FlexiMMT_page/) | [**GitHub**](https://github.com/Ethan-Li123/FlexiMMT)
|
| 14 |
+
|
| 15 |
+
## Data Preparation
|
| 16 |
+
|
| 17 |
+
You can download the data used in the paper by cloning this repository:
|
| 18 |
+
|
| 19 |
+
```bash
|
| 20 |
+
git lfs install
|
| 21 |
+
git clone https://huggingface.co/datasets/llyyzzz/FlexiMMT ./benchmark_new
|
| 22 |
+
```
|
| 23 |
+
|
| 24 |
+
### Data Structure
|
| 25 |
+
|
| 26 |
+
The data structure is organized as follows:
|
| 27 |
+
|
| 28 |
+
```
|
| 29 |
+
|-- benchmark_new
|
| 30 |
+
|-- captions_train
|
| 31 |
+
|-- animal
|
| 32 |
+
|-- bear
|
| 33 |
+
|-- crop.csv
|
| 34 |
+
|-- val_image.csv
|
| 35 |
+
|-- ...
|
| 36 |
+
|-- human
|
| 37 |
+
|-- chest
|
| 38 |
+
|-- crop.csv
|
| 39 |
+
|-- val_image.csv
|
| 40 |
+
|-- ...
|
| 41 |
+
|-- captions_inf
|
| 42 |
+
|-- val_images.csv
|
| 43 |
+
|-- reference_videos
|
| 44 |
+
|-- animal
|
| 45 |
+
|-- bear_crop/
|
| 46 |
+
|-- ...
|
| 47 |
+
|-- human
|
| 48 |
+
|-- chest_crop/
|
| 49 |
+
|-- ...
|
| 50 |
+
|-- reference_video_masks_train
|
| 51 |
+
|-- animal
|
| 52 |
+
|-- bear_crop/
|
| 53 |
+
|-- ...
|
| 54 |
+
|-- human
|
| 55 |
+
|-- ...
|
| 56 |
+
|-- reference_video_masks_eval
|
| 57 |
+
|-- ...
|
| 58 |
+
|-- target_images
|
| 59 |
+
|-- 0_bear1+movie_man_1.png
|
| 60 |
+
|-- ...
|
| 61 |
+
|-- target_masks
|
| 62 |
+
|-- {image_name}+{action}/
|
| 63 |
+
|-- ...
|
| 64 |
+
```
|
| 65 |
+
|
| 66 |
+
## Citation
|
| 67 |
+
|
| 68 |
+
```bibtex
|
| 69 |
+
@article{li2026letimagemotion,
|
| 70 |
+
title={Let Your Image Move with Your Motion! -- Implicit Multi-Object Multi-Motion Transfer},
|
| 71 |
+
author={Yuze Li and Dong Gong and Xiao Cao and Junchao Yuan and Dongsheng Li and Lei Zhou and Yun Sing Koh and Cheng Yan and Xinyu Zhang},
|
| 72 |
+
year={2026},
|
| 73 |
+
eprint={2603.01000},
|
| 74 |
+
archivePrefix={arXiv},
|
| 75 |
+
primaryClass={cs.CV},
|
| 76 |
+
url={https://arxiv.org/abs/2603.01000},
|
| 77 |
+
}
|
| 78 |
+
```
|