2018年用于杂草多光谱图像数据集

杂草地图数据集由发布weedNet数据集的作者创建。该数据集体量较大，有可能是目前公开数据库中最大的用于甜菜杂草分割和制图的多光谱航空数据集。数据采集通过两架无人机分别配备四通道和五通道多光谱相机，用于从甜菜田10m高度处采集图像。数据集包括八组高分辨率正射镶嵌图，带有针对作物、杂草和背景的像素级注释，以及总共 10,196个标题图像。标题图像是指从上述正射镶嵌图中以滑动窗口方式裁剪的小图像块或瓦片。该数据集提供了机器学习算法的新基准，可用于生成基于大规模正射镶嵌图的杂草映射研究。https://projects.asl.ethz.ch/datasets/doku.php?id=weedmap:remotesensing2018weedmap#dataset_summary

The WeedMap dataset was created by the authors who published the weedNet dataset. With a large scale, this dataset is likely the largest publicly available multispectral aerial dataset for sugar beet weed segmentation and mapping to date. Data was collected using two unmanned aerial vehicles (UAVs) equipped with four-channel and five-channel multispectral cameras respectively, to capture images at a height of 10 meters above sugar beet fields. The dataset includes eight sets of high-resolution orthomosaics, with pixel-level annotations for crops, weeds and background, as well as a total of 10,196 tile images. A tile image refers to a small image patch or tile cropped from the aforementioned orthomosaics via a sliding window method. This dataset provides a novel benchmark for machine learning algorithms to support weed mapping research based on large-scale orthomosaics. https://projects.asl.ethz.ch/datasets/doku.php?id=weedmap:remotesensing2018weedmap#dataset_summary