2019年用于农业中语义和层次物种分析的草和三叶草图像数据集

草和三叶草数据集是多样化图像分割和生物量数据集，旨在支持对严重遮挡的混合作物进行稳健图像分析以进行精确管理。这些图像包含密集的草和三叶草混合种群，有严重遮挡和各种杂草出现。数据集通过三个带有数码相机的地面采集平台收集。数据集分为训练数据集集和测试数据集。训练数据集由8000张带有像素注释的合成图像、31600张未标记图像和另外152张带有像素注释的图像组成。植物冠层生物量组成信息均为jpg格式。合成图像根据原始图像中不同物种和土壤背景的随机整合生成，以便创建大量带注释的图像，减少重复工作量。测试集由15张人工标注图像和283张带有生物量信息的图像组成。该数据集是第一个支持图像分割和生物量组成预测任务的数据集。https://vision.eng.au.dk/grass-clover-dataset

The Grass and Clover Dataset is a diversified image segmentation and biomass dataset designed to support robust image analysis of heavily occluded mixed crops for precise management. These images feature dense mixed populations of grass and clover, with severe occlusions and the presence of various weeds. The dataset was collected via three ground-based acquisition platforms equipped with digital cameras. The dataset is split into training and test subsets. The training subset consists of 8000 synthetic images with pixel-level annotations, 31600 unlabeled images, and an additional 152 images with pixel-level annotations. All plant canopy biomass composition information is stored in JPEG format. Synthetic images are generated through random integration of different species and soil backgrounds from original images, enabling the creation of large quantities of annotated images while reducing repetitive workload. The test subset comprises 15 manually annotated images and 283 images with biomass information. This dataset is the first of its kind to support both image segmentation and biomass composition prediction tasks. https://vision.eng.au.dk/grass-clover-dataset