应用于农业场景视觉解析任务的番茄数据集

该数据集由两部分组成，分别为合成部分和经验部分。其中合成部分使用Wageningen大学等人提出的数据合成方法，生成3250张合成番茄图像以及对应的像素级别语义分割标签图；经验部分由RGB相机拍摄的750张单目图像和400张双目图像构成，并人工对它们部分进行了包括实例分割、目标检测等在内的精细标注。研究旨在从多个角度丰富该数据集，包括数据集的大小、标注信息的多维度、场景的复杂性等方面，为今后利用计算机视觉技术解决农业领域问题提供数据基础

This dataset comprises two subsets: the synthetic subset and the real-world empirical subset. The synthetic subset adopts the data synthesis method proposed by Wageningen University et al. to generate 3250 synthetic tomato images along with their corresponding pixel-level semantic segmentation label maps. The real-world empirical subset consists of 750 monocular images and 400 stereo images captured by RGB cameras, and some of these images have been manually annotated with fine-grained labels including instance segmentation, object detection and other relevant computer vision tasks. This study aims to enrich this dataset from multiple dimensions, including dataset scale, multi-dimensional annotation information and scene complexity, so as to provide a solid data foundation for future computer vision-based solutions to agricultural problems.