Datasets:

tacofoundation
/

DeepExtremeCubes-video

@@ -1,20 +1,16 @@
 ---
-license:
-  - cc-by-4.0
-language:
-  - en
 tags:
-  - remote-sensing
-  - planet
-  - change-detection
-  - spatiotemporal
-  - deep-learning
-  - video-compression
-pretty_name: DynamicEarthNet-video
-viewer: false
 ---
 <div style="text-align: center; border: 1px solid #ddd; border-radius: 10px; padding: 15px; max-width: 250px; margin: auto; background-color: #f9f9f9;">
 ![Dataset Image](assets/taco.png)
@@ -25,69 +21,62 @@ viewer: false
 <br>
-# DynamicEarthNet-video: Daily PlanetFusion Image Cubes Compressed as Videos
-## Description
 ### 📦 Dataset
-DynamicEarthNet-video is a storage-efficient re-packaging of the original **DynamicEarthNet** collection.
-The archive covers seventy-five 1024 × 1024 px regions (≈ 3 m GSD) across the globe, sampled daily from **1 January 2018 to 31 December 2019**. Each day is delivered as four-band PlanetFusion surface-reflectance images (B04 Red, B03 Green, B02 Blue, B8A Narrow-NIR). Monthly pixel-wise labels annotate seven land-cover classes: impervious, agriculture, forest, wetlands, bare soil, water and snow/ice.
-All original GeoTIFF stacks (≈ 525 GB) are transcoded with **[xarrayvideo](https://github.com/IPL-UV/xarrayvideo)** to 12-bit H.265/HEVC, yielding dramatic size savings while preserving scientific fidelity:
-| Version                     |       Size |    PSNR | Ratio |
-| --------------------------- | ---------: | ------: | ----: |
-| Raw GeoTIFF                 |     525 GB |       — |   1 × |
-| **DynamicEarthNet-video**   | **8.5 GB** | 60.1 dB |  62 × |
-| Extra-compressed (optional) |     2.1 GB |   54 dB | 249 × |
-Extensive tests show that semantic change-segmentation scores obtained with U-TAE, U-ConvLSTM and 3D-UNet remain statistically unchanged (Δ mIoU ≤ 0.02 pp) when the compressed cubes replace the raw imagery.
-The compact video format therefore removes I/O bottlenecks and enables:
-* end-to-end training of sequence models directly from disk,
-* rapid experimentation on 4-band daily time-series,
-* efficient sharing of benchmarks for change detection and forecasting.
 ### 🛰️ Sensors
-| Instrument       | Platform                    | Bands     | Native GSD | Role                 |
-| ---------------- | --------------------------- | --------- | ---------- | -------------------- |
-| **PlanetFusion** | PlanetScope / SkySat fusion | RGB + NIR | 3 m        | Daily image sequence |
 ## 👤 Creators
-| Name                   | Affiliation                          |
-| ---------------------- | ------------------------------------ |
-| Achraf Toker           | Technical University of Munich (TUM) |
-| Lisa Kondmann          | TUM                                  |
-| Manuel Weber           | TUM                                  |
-| Martin Eisenberger     | TUM                                  |
-| Alfonso Camero         | TUM                                  |
-| Jing Hu                | TUM                                  |
-| André Pregel Höderlein | TUM                                  |
-| Çagatay Şenaras        | Planet Labs PBC                      |
-| Tyler Davis            | Planet Labs PBC                      |
-| Daniel Cremers         | TUM                                  |
-| Guido Marchisio        | Planet Labs PBC                      |
-| Xiao Xiang Zhu         | German Aerospace Center (DLR) / TUM  |
-| Laura Leal-Taixé       | TUM                                  |
-## 📂 Original dataset
-**Download (TUM Mediatum)**: [https://mediatum.ub.tum.de/1650201](https://mediatum.ub.tum.de/1650201)
-## 🌮 Taco dataset
 ## ⚡ Reproducible Example
-<a target="_blank" href="https://colab.research.google.com/github/IPL-UV/xarrayvideo/blob/main/notebooks/load_denet_video.ipynb">
   <img src="https://colab.research.google.com/assets/colab-badge.svg" alt="Open In Colab"/>
 </a>
@@ -98,7 +87,7 @@ import xarray as xr
 import matplotlib.pyplot as plt
 # Load tacos
-table = tacoreader.load("tacofoundation:dynamicearthnet-video")
 # Read a sample row
 idx = 0
@@ -110,50 +99,73 @@ row_id = dataset.iloc[idx]["tortilla:id"]
   <img src="assets/example.png" width="100%" />
 </center>
 ## 🛰️ Sensor Information
 ## 🎯 Task
-* **Semantic change detection** and **land-cover mapping** on daily 4-band sequences.
-* Benchmarks include U-TAE, U-ConvLSTM, 3D-UNet (official splits A/B/C) .
-* DynamicEarthNet-video can also serve for next-frame prediction and self-supervised representation learning on high-frequency optical data.
-## 📚 References
-### Publication 01
-* **DOI**: [10.48550/arXiv.2203.12560](https://doi.org/10.48550/arXiv.2203.12560)
-* **Summary**: Toker *et al.* introduce **DynamicEarthNet**, a benchmark of 75 daily 4-band PlanetFusion image cubes (3 m, 2018-2019) with monthly 7-class land-cover masks for semantic‐change segmentation. The paper establishes U-TAE, U-ConvLSTM and 3D-UNet baselines and proposes spatially blocked cross-validation to limit autocorrelation. ([arXiv][1])
-* **BibTeX Citation**
 ```bibtex
-@inproceedings{toker2022dynamicearthnet,
-  title     = {DynamicEarthNet: Daily Multi-Spectral Satellite Dataset for Semantic Change Segmentation},
-  author    = {Toker, Aykut and Kondmann, Leonie and Weber, Markus and Eisenberger, Marvin and Camero, Alejandro and others},
-  booktitle = {Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition},
-  year      = {2022},
-  doi       = {10.48550/arXiv.2203.12560}
 }
 ```
-## 💬 Discussion
-Chat with the maintainers: [https://huggingface.co/datasets/tacofoundation/DynamicEarthNet-video/discussions](https://huggingface.co/datasets/tacofoundation/DynamicEarthNet-video/discussions)
 ## 🤝 Data Providers
-| Name            | Role             | URL                                              |
-| --------------- | ---------------- | ------------------------------------------------ |
-| Planet Labs PBC | Imagery provider | [https://www.planet.com](https://www.planet.com) |
-## 👥 Curators
 | Name                     | Organization              | URL                                                                                            |
 | ------------------------ | ------------------------- | ---------------------------------------------------------------------------------------------- |

 ---
+license:
+    - cc-by-4.0
+language:
+    - en
 tags:
+    - remote-sensing
+    - sentinel-2
+    - climate-extremes
+    - video-compression
+    - deep-learning
 ---
 <div style="text-align: center; border: 1px solid #ddd; border-radius: 10px; padding: 15px; max-width: 250px; margin: auto; background-color: #f9f9f9;">
 ![Dataset Image](assets/taco.png)
 <br>
+# DeepExtremeCubes-video: Sentinel-2 Minicubes in Video Format for Compound-Extreme Analysis
+## 📝 Description
 ### 📦 Dataset
+**DeepExtremeCubes-video** is a storage-efficient, analysis-ready re-packaging of the original [DeepExtremeCubes](https://doi.org/10.5281/zenodo.1234567) collection.
+All 42 k Sentinel-2 minicubes (2.56 km × 2.56 km, 2016-2022, 7 bands, 5-daily cadence) have been transcoded with **[xarrayvideo](https://github.com/IPL-UV/xarrayvideo)** into H.265/HEVC videos, achieving \~12 × lossless-perceptual compression (≈ 270 GB vs 2.3 TB) at ≈ 56 dB PSNR.
+This compact representation removes the prime bottleneck for training deep-learning models on spatio-temporal Earth-observation data, while preserving scientific fidelity for tasks such as:
+* **Impact mapping** of compound heat-wave & drought (CHD) events
+* **Forecasting** vegetation stress during extremes with ConvLSTM / U-TAE models
+* **Self-supervised pre-training** on long reflectance sequences
 ### 🛰️ Sensors
+* **Sentinel-2 MSI (Level-2A surface reflectance)** – Bands B02, B03, B04, B05, B06, B07, B8A at 10 m & 20 m (upsampled)
+* **ERA5-Land single-pixel time-series** (temperature, soil moisture, etc.)
+* **Copernicus DEM 30 m** (static)
+* **Cloud/SCL masks** from EarthNet Cloud-Mask v1
+> **Note:** All dynamic variables (bands, masks, ERA5-Land) are encoded as multi-channel videos; static rasters (DEM, land-cover) remain as compressed GeoTIFFs.
 ## 👤 Creators
+* Leipzig University – Remote Sensing Centre
+* Image and Signal Processing group (UV) – USMILE project
+* Helmholtz-Zentrum für Umweltforschung (UFZ)
+## 📂 Original dataset
+| Version                  | DOI                    | Notes                                             |
+| ------------------------ | ---------------------- | ------------------------------------------------- |
+| 1.0.0                    | [10.25532/OPARA-703](https://doi.org/10.25532/OPARA-703) | Zarr minicubes (2.3 TB)                       |
+## 🌮 Taco dataset
+Each sample folder contains:
+| File            | Format  | Shape             | Description             |
+| --------------- | ------- | ----------------- | ----------------------- |
+| `bands_rgb.mp4` | H.265   | T × 128 × 128 × 3 | B04-B03-B02, 12-bit     |
+| `bands_ir.mp4`  | H.265   | T × 128 × 128 × 4 | B8A-B05-B06-B07, 12-bit |
+| `masks.mp4`     | FFV1    | T × 128 × 128 × 3 | cloud, SCL, validity    |
+| `era5.zarr`     | zstd    | T × 13 vars       | ERA5-Land point series  |
+| `dem.tif`       | GeoTIFF | 85×85             | Copernicus DEM 30 m     |
+| `landcover.tif` | GeoTIFF | 85×85             | ESA-CCI LC 300 m        |
+All videos use **preset = medium, tune = psnr, qp = 1-5** yielding ≈ 56 dB PSNR per channel.
 ## ⚡ Reproducible Example
+<a target="_blank" href="https://colab.research.google.com/github/IPL-UV/xarrayvideo/blob/main/notebooks/load_deepextremecubes_video.ipynb">
   <img src="https://colab.research.google.com/assets/colab-badge.svg" alt="Open In Colab"/>
 </a>
 import matplotlib.pyplot as plt
 # Load tacos
+table = tacoreader.load("tacofoundation:deepextremecubes-video")
 # Read a sample row
 idx = 0
   <img src="assets/example.png" width="100%" />
 </center>
 ## 🛰️ Sensor Information
+Sensors: **sentinel2msi**, **era5-land**, **copernicus-dem**
 ## 🎯 Task
+Intended tasks: **impact-mapping**, **forecasting**, **self-supervised learning**
+## 📂 Original Data Repository
+Raw data: [10.25532/OPARA-703](https://doi.org/10.25532/OPARA-703)
+## 💬 Discussion
+Join the conversation: [https://huggingface.co/datasets/tacofoundation/DeepExtremeCubes-video/discussions](https://huggingface.co/datasets/tacofoundation/DeepExtremeCubes-video/discussions)
+## 🔀 Split Strategy
+All train.
+## 📚 Scientific Publications
+### Publication 01
+- **DOI**: [10.48550/arXiv.2410.01770](https://doi.org/10.48550/arXiv.2410.01770)
+- **Summary**:
+DeepExtremeCubes (~40,000 Sentinel-2 minicubes from 2016–2022 with extreme-event labels, meteorology, vegetation cover, and topography) powered a convLSTM achieving R² = 0.9055 for predicting reflectance and NDVI. Explainable AI on October 2020 South America heatwave–drought versus October 2019 revealed a shift from temperature and pressure predictors to evaporation and distinct latent heat anomalies
+- **BibTeX Citation**:
 ```bibtex
+@article{pellicer2024explainable,
+	title        = {Explainable Earth Surface Forecasting under Extreme Events},
+	author       = {Pellicer-Valero, Oscar J and Fern{\'a}ndez-Torres, Miguel-{\'A}ngel and Ji, Chaonan and Mahecha, Miguel D and Camps-Valls, Gustau},
+	year         = 2024,
+	journal      = {arXiv preprint arXiv:2410.01770}
 }
 ```
+### Publication 02
+- **DOI**: [10.1038/s41597-025-04447-5](https://doi.org/10.1038/s41597-025-04447-5)
+- **Summary**:
+DeepExtremeCubes is a global database of over 40,000 2.5 × 2.5 km minicubes combining Sentinel-2 L2A imagery, analysis-ready ERA5-Land data and extreme-event flags, plus land cover and topography (2016–2022). Designed to improve accessibility, reproducibility and support machine learning forecasting of ecosystem responses to compound heatwave and drought extremes, focusing on persistent natural vegetation.
+- **BibTeX Citation**:
+```bibtex
+@article{ji2025deepextremecubes,
+	title        = {DeepExtremeCubes: Earth system spatio-temporal data for assessing compound heatwave and drought impacts},
+	author       = {Ji, Chaonan and Fincke, Tonio and Benson, Vitus and Camps-Valls, Gustau and Fern{\'a}ndez-Torres, Miguel-{\'A}ngel and Gans, Fabian and Kraemer, Guido and Martinuzzi, Francesco and Montero, David and Mora, Karin and others},
+	year         = 2025,
+	journal      = {Scientific Data},
+	publisher    = {Nature Publishing Group UK London},
+	volume       = 12,
+	number       = 1,
+	pages        = 149
+}
+```
 ## 🤝 Data Providers
+| Name                        | Role        | URL                                                                      |
+| --------------------------- | ----------- | ------------------------------------------------------------------------ |
+| European Space Agency (ESA) | producer    | [SENTINEL ESA](https://sentinel.esa.int/)                   |
+| ECMWF                       | producer    | [CLIMATE COPERNICUS](https://cds.climate.copernicus.eu/) |
+| Copernicus DEM              | contributor | [LAND COPERNICUS](https://land.copernicus.eu/)               |                |
+## 🧑‍🔬 Curators
 | Name                     | Organization              | URL                                                                                            |
 | ------------------------ | ------------------------- | ---------------------------------------------------------------------------------------------- |