EPFL-ECEO/EcoWikiRS

EcoWikiRS数据集由图像、物种列表和维基百科句子组成的三元组构成，包含91k高分辨率航空图像（50cm，RGB波段）、来自2745种不同物种的众包物种观察数据（从GBIF收集，并在地理位置上与航空图像对应）以及103K个描述观察物种栖息地的句子（从相应的维基百科文章中提取）。数据集分为原始原始数据和Hugging Face处理版本，具有唯一标识符、栖息地类别标签、物种列表、图像和物种文本数据等特征。

The EcoWikiRS dataset consists of triplets of images, species lists, and Wikipedia sentences, including 91k high-resolution aerial images (50cm, RGB bands), crowd-sourced species observations from 2745 different species (collected from GBIF and geolocated within the footprint of the aerial image), and 103K different sentences describing the habitat of the observed species (extracted from the corresponding Wikipedia article). The dataset is divided into original raw data and a Hugging Face processed version, featuring unique identifiers, habitat class labels, species lists, images, and species text data.