YuxiangZhang-BIT/UniTS-Datasets-ckpt

---
license: cc-by-sa-4.0
size_categories:
- n>1T
pretty_name: UniTS_TS-S12&TS-S12CR
viewer: false
tags:
- Earth observation
- Multimodal
- Time series
- Time Series Reconstruction
- Time Series Cloud Removal
- Time Series Forecasting
---

# UniTS_TS-S12&TS-S12CR
[[Paper 📰]](https://arxiv.org/abs/2512.04461) [[Project Page ✨]](https://yuxiangzhang-bit.github.io/UniTS-website/)

We construct two high-quality multimodal time-series datasets, namely TS-S12 and TS-S12CR. Among them, TS-S12 and TS-S12CR contain Sentinel-1 with 2 channels (VV and VH) and Sentinel-2 imagery with 10 spectral bands (excluding B1 Aerosols, B9 Water Vapor, and B10 Cirrus) from 14,973 and 12,126 ROIs around the world, respectively.
<img src="assets/dataset.png" style="max-width: 100%; height: auto;">
<img src="assets/dataset-tab.jpg" style="max-width: 100%; height: auto;">
- TS-S12 provides aligned sample pairs of **Sentinel-1**, **cloud-free Sentinel-2** and **all annual cloud masks** for time series reconstruction and forecasting tasks.
- TS-S12CR offers aligned triplets samples of **Sentinel-1**, **cloud-covered Sentinel-2**, and **cloud-free Sentinel-2** specifically designed for time series cloud removal task. TS-S12CR provides an extreme scenario with an **average cloud coverage of 84.02%**, serving as an important benchmark for developing robust time series cloud removal methods.

---

## Dataset organisation

- TS-S12 dataset directory (**1.5T**)
```bash
TS-S12/
├── data_s1_s2/ # Sentinel-1 & 2 for each ROI, 256×256 patch size
│ ├── roi851_s1_f39.h5 # roi{ROI}_s1_f{sequence lengths}.h5
│ ├── roi851_s2_f39.h5 # roi{ROI}_s2_f{sequence lengths}.h5 | cloud-free Sentinel-2
│ └── ...
├── cldmsk/ # All annual cloud masks for each ROI
│ ├── roi851_cldmsk.h5 # roi{ROI}_cldmsk.h5
│ └── ...
└── splits/
├── train.txt
└── test.txt
```
- TS-S12CR dataset directory (**767G**)
```bash
TS-S12CR/
├── data_s1_s2/ # Sentinel-1 & 2 for each ROI, 256×256 patch size
│ ├── roi2434_s1_f46.h5 # roi{ROI}_s1_f{sequence lengths}.h5
│ ├── roi2434_s2_f46_wcld.h5 # roi{ROI}_s2_f{sequence lengths}_wcld.h5 | cloud-cover Sentinel-2
│ ├── roi2434_s2_f46.h5 # roi{ROI}_s2_f{sequence lengths}.h5 | cloud-free Sentinel-2
│ └── ...
└── splits/
├── train.txt
└── test.txt
```
- Land cover map directory (**45G**)
```bash
land_cover_map/
├── roi851_s2_f39.h5
├── roi2434_s2_f46.h5
└── ...
```
---

## Download

You can download the dataset with the Hugging Face CLI tool.

```shell
pip install huggingface_hub
hf download YuxiangZhang-BIT/UniTS-Datasets-ckpt --repo-type dataset --local-dir data/UniTS-Datasets
```

If you like to download only a subset of the data, you can specify it with `--include`.
```
# Only download val data
hf download YuxiangZhang-BIT/UniTS-Datasets-ckpt --repo-type dataset --include "TS-S12/*" --local-dir data/UniTS-Datasets

```

---

## Citation

If you use TS-S12&TS-S12CR datasets, please cite:

```bibtex
@article{zhang2025unitsunifiedtimeseries,
title={UniTS: Unified Time Series Generative Model for Remote Sensing},
author={Yuxiang Zhang and Shunlin Liang and Wenyuan Li and Han Ma and Jianglei Xu and Yichuan Ma and Jiangwei Xie and Wei Li and Mengmeng Zhang and Ran Tao and Xiang-Gen Xia},
year={2025},
eprint={2512.04461},
archivePrefix={arXiv},
url={https://arxiv.org/abs/2512.04461},
}
```