DynamicEarthNet: Daily Multi-Spectral Satellite Dataset for Semantic Change Segmentation - validation and test label

This is the dataset repository of the paper: Toker, A.*, Kondmann, L.*, Weber, M., Eisenberger, M., Camero, A., Hu, J., Pregel Hoderlein, A., Senaras, C., Davis, T., Cremers, D., Marchisio, G°., Zhu, X.X.°, Leal-Taixé, L.°: DynamicEarthNet: Daily multi-spectral satellite dataset for semantic change segmentation. In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (2022)
<br>
This is the validation and test label of the DynamicEarthNet dataset. The dataset (see below link to version 1) contains daily Planet Fusion imagery with monthly land cover classes for 75 areas across the globe over two years. The seven land cover classes were manually annotated in a temporally consistent way. Sentinel 2 imagery is also provided. This dataset is the first large-scale multi-class and multi-temporal change detection benchmark and we hope it will foster a new wave of multi-temporal research in Earth Observation as well as Computer Vision. A detailed description is available in the paper.

本数据集仓库对应论文：Toker, A.*, Kondmann, L.*, Weber, M., Eisenberger, M., Camero, A., Hu, J., Pregel Hoderlein, A., Senaras, C., Davis, T., Cremers, D., Marchisio, G°., Zhu, X.X.°, Leal-Taixé, L.°：DynamicEarthNet：用于语义变化分割的每日多光谱卫星数据集。收录于《IEEE/CVF计算机视觉与模式识别会议论文集》（2022）<br>本文件为DynamicEarthNet数据集的验证集与测试集标签。该数据集（详见下文版本1链接）包含全球75个区域两年间的每日Planet Fusion影像，附带逐月土地覆盖类别标注。七种土地覆盖类别均按照时间一致性原则完成人工标注。此外还提供了哨兵2号（Sentinel 2）影像。本数据集是首个大规模多类别多时相变化检测基准数据集，我们期望其能够催生地球观测（Earth Observation）与计算机视觉领域内多时相研究的新热潮。详细说明可参阅上述论文。

搜集汇总

背景与挑战

背景概述

DynamicEarthNet验证和测试标签数据集是一个用于语义变化分割的多光谱卫星数据集，包含全球75个地区两年内每日的Planet Fusion影像和每月土地覆盖类别，以及Sentinel 2影像，所有数据均经过人工标注并保持时间一致性。该数据集旨在推动地球观测和计算机视觉领域多时相研究的发展。