Datasets:
Upload
Browse files- README.md +162 -0
- assets/taco.png +3 -0
README.md
ADDED
@@ -0,0 +1,162 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
---
|
2 |
+
license:
|
3 |
+
- cc-by-4.0
|
4 |
+
language:
|
5 |
+
- en
|
6 |
+
tags:
|
7 |
+
- remote-sensing
|
8 |
+
- planet
|
9 |
+
- change-detection
|
10 |
+
- spatiotemporal
|
11 |
+
- deep-learning
|
12 |
+
- video-compression
|
13 |
+
pretty_name: DynamicEarthNet-video
|
14 |
+
viewer: false
|
15 |
+
---
|
16 |
+
|
17 |
+
|
18 |
+
<div style="text-align: center; border: 1px solid #ddd; border-radius: 10px; padding: 15px; max-width: 250px; margin: auto; background-color: #f9f9f9;">
|
19 |
+
|
20 |
+

|
21 |
+
|
22 |
+
<b><p>This dataset follows the TACO specification.</p></b>
|
23 |
+
</div>
|
24 |
+
|
25 |
+
<br>
|
26 |
+
|
27 |
+
|
28 |
+
# DynamicEarthNet-video: Daily PlanetFusion Image Cubes Compressed as Videos
|
29 |
+
|
30 |
+
## Description
|
31 |
+
|
32 |
+
|
33 |
+
### 📦 Dataset
|
34 |
+
|
35 |
+
DynamicEarthNet-video is a storage-efficient re-packaging of the original **DynamicEarthNet** collection.
|
36 |
+
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.
|
37 |
+
|
38 |
+
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:
|
39 |
+
|
40 |
+
| Version | Size | PSNR | Ratio |
|
41 |
+
| --------------------------- | ---------: | ------: | ----: |
|
42 |
+
| Raw GeoTIFF | 525 GB | — | 1 × |
|
43 |
+
| **DynamicEarthNet-video** | **8.5 GB** | 60.1 dB | 62 × |
|
44 |
+
| Extra-compressed (optional) | 2.1 GB | 54 dB | 249 × |
|
45 |
+
|
46 |
+
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.
|
47 |
+
|
48 |
+
The compact video format therefore removes I/O bottlenecks and enables:
|
49 |
+
|
50 |
+
* end-to-end training of sequence models directly from disk,
|
51 |
+
* rapid experimentation on 4-band daily time-series,
|
52 |
+
* efficient sharing of benchmarks for change detection and forecasting.
|
53 |
+
### 🛰️ Sensors
|
54 |
+
|
55 |
+
| Instrument | Platform | Bands | Native GSD | Role |
|
56 |
+
| ---------------- | --------------------------- | --------- | ---------- | -------------------- |
|
57 |
+
| **PlanetFusion** | PlanetScope / SkySat fusion | RGB + NIR | 3 m | Daily image sequence |
|
58 |
+
|
59 |
+
|
60 |
+
## 👤 Creators
|
61 |
+
|
62 |
+
|
63 |
+
| Name | Affiliation |
|
64 |
+
| ---------------------- | ------------------------------------ |
|
65 |
+
| Achraf Toker | Technical University of Munich (TUM) |
|
66 |
+
| Lisa Kondmann | TUM |
|
67 |
+
| Manuel Weber | TUM |
|
68 |
+
| Martin Eisenberger | TUM |
|
69 |
+
| Alfonso Camero | TUM |
|
70 |
+
| Jing Hu | TUM |
|
71 |
+
| André Pregel Höderlein | TUM |
|
72 |
+
| Çagatay Şenaras | Planet Labs PBC |
|
73 |
+
| Tyler Davis | Planet Labs PBC |
|
74 |
+
| Daniel Cremers | TUM |
|
75 |
+
| Guido Marchisio | Planet Labs PBC |
|
76 |
+
| Xiao Xiang Zhu | German Aerospace Center (DLR) / TUM |
|
77 |
+
| Laura Leal-Taixé | TUM |
|
78 |
+
|
79 |
+
|
80 |
+
## 📂 Original dataset
|
81 |
+
|
82 |
+
**Download (TUM Mediatum)**: [https://mediatum.ub.tum.de/1650201](https://mediatum.ub.tum.de/1650201)
|
83 |
+
|
84 |
+
|
85 |
+
|
86 |
+
## 🌮 Taco dataset
|
87 |
+
|
88 |
+
## ⚡ Reproducible Example
|
89 |
+
|
90 |
+
<a target="_blank" href="https://colab.research.google.com/drive/1V3kfJmbWJRVncQwbdqLKgDp4-adMVy4N?usp=sharing">
|
91 |
+
<img src="https://colab.research.google.com/assets/colab-badge.svg" alt="Open In Colab"/>
|
92 |
+
</a>
|
93 |
+
|
94 |
+
```python
|
95 |
+
import tacoreader
|
96 |
+
import xarrayvideo as xav
|
97 |
+
import xarray as xr
|
98 |
+
import matplotlib.pyplot as plt
|
99 |
+
|
100 |
+
# Load tacos
|
101 |
+
table = tacoreader.load("tacofoundation:dynamicearthnet-video")
|
102 |
+
|
103 |
+
# Read a sample row
|
104 |
+
idx = 0
|
105 |
+
row = dataset.read(idx)
|
106 |
+
row_id = dataset.iloc[idx]["tortilla:id"]
|
107 |
+
```
|
108 |
+
|
109 |
+
<center>
|
110 |
+
<img src="assets/example.png" width="100%" />
|
111 |
+
</center>
|
112 |
+
|
113 |
+
|
114 |
+
## 🛰️ Sensor Information
|
115 |
+
|
116 |
+
Sensors: **planet**
|
117 |
+
|
118 |
+
## 🎯 Task
|
119 |
+
|
120 |
+
|
121 |
+
* **Semantic change detection** and **land-cover mapping** on daily 4-band sequences.
|
122 |
+
* Benchmarks include U-TAE, U-ConvLSTM, 3D-UNet (official splits A/B/C) .
|
123 |
+
* DynamicEarthNet-video can also serve for next-frame prediction and self-supervised representation learning on high-frequency optical data.
|
124 |
+
|
125 |
+
## 📚 References
|
126 |
+
|
127 |
+
|
128 |
+
### Publication 01
|
129 |
+
|
130 |
+
* **DOI**: [10.48550/arXiv.2203.12560](https://doi.org/10.48550/arXiv.2203.12560)
|
131 |
+
* **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])
|
132 |
+
* **BibTeX Citation**
|
133 |
+
|
134 |
+
```bibtex
|
135 |
+
@inproceedings{toker2022dynamicearthnet,
|
136 |
+
title = {DynamicEarthNet: Daily Multi-Spectral Satellite Dataset for Semantic Change Segmentation},
|
137 |
+
author = {Toker, Aykut and Kondmann, Leonie and Weber, Markus and Eisenberger, Marvin and Camero, Alejandro and others},
|
138 |
+
booktitle = {Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition},
|
139 |
+
year = {2022},
|
140 |
+
doi = {10.48550/arXiv.2203.12560}
|
141 |
+
}
|
142 |
+
```
|
143 |
+
|
144 |
+
|
145 |
+
## 💬 Discussion
|
146 |
+
|
147 |
+
Chat with the maintainers: [https://huggingface.co/datasets/tacofoundation/DynamicEarthNet-video/discussions](https://huggingface.co/datasets/tacofoundation/DynamicEarthNet-video/discussions)
|
148 |
+
|
149 |
+
|
150 |
+
## 🤝 Data Providers
|
151 |
+
|
152 |
+
| Name | Role | URL |
|
153 |
+
| --------------- | ---------------- | ------------------------------------------------ |
|
154 |
+
| Planet Labs PBC | Imagery provider | [https://www.planet.com](https://www.planet.com) |
|
155 |
+
|
156 |
+
## 👥 Curators
|
157 |
+
|
158 |
+
| Name | Organization | URL |
|
159 |
+
| ------------------------ | ------------------------- | ---------------------------------------------------------------------------------------------- |
|
160 |
+
| Oscar J. Pellicer-Valero | Image Signal Processing (ISP) | [Google Scholar](https://scholar.google.com/citations?user=CCFJshwAAAAJ&hl=en) |
|
161 |
+
| Cesar Aybar | Image Signal Processing (ISP) | [Google Scholar](https://scholar.google.es/citations?user=rfF51ocAAAAJ&hl=es) |
|
162 |
+
| Julio Contreras | Image Signal Processing (ISP) | [GitHub](https://github.com/JulioContrerasH) |
|
assets/taco.png
ADDED
![]() |
Git LFS Details
|