2020年开放植物表型图像数据库（OPPD）

开放植物表型数据库OPPD是从47 种杂草幼苗中收集数据形成的数据集。这些植物在三种不同生长条件（即理想、干旱和自然）下的半田间受控环境中培养，以确保植物视觉外观的高度种内变化。研究者在四个试验生长季节进行数据收集，并从幼苗出苗到多达6-8片叶子阶段起，对每个试验的植物进行时间跟踪。生成的数据集包含 7,590 张jpg 格式的 红绿蓝 图像，代表64,292株植物。每个图像都标注了相应杂草种类的标签和植物的边界框，数据加工工程通过基于机器学习的标注工具 RoboWeedMaPS(https://vision.eng.au.dk/roboweedmaps/)完成，并进行手动校正。该数据集可用于评估植物分类和实例检测任务。https://gitlab.au.dk/AUENG-Vision/OPPD/-/tree/master/

The Open Plant Phenotyping Database (OPPD) is a dataset compiled from data collected from 47 species of weed seedlings. These plants were grown in a semi-field controlled environment under three distinct growth conditions—optimal, drought, and natural—to ensure high intraspecific variation in their visual appearance. Researchers carried out data collection across four experimental growing seasons, and conducted temporal tracking of plants in each trial from seedling emergence until the plants reached the 6-to-8-leaf stage. The resultant dataset comprises 7,590 red-green-blue (RGB) images in JPG format, representing a total of 64,292 individual plants. Each image is annotated with the corresponding weed species label and a bounding box for the plant. The data processing was completed using the machine learning-powered annotation tool RoboWeedMaPS (https://vision.eng.au.dk/roboweedmaps/), with subsequent manual verification and correction. This dataset can be utilized to evaluate performance on plant classification and instance detection tasks. The dataset is publicly accessible at https://gitlab.au.dk/AUENG-Vision/OPPD/-/tree/master/

搜集汇总

背景与挑战

背景概述

2020年开放植物表型图像数据库（OPPD）是一个包含47种杂草幼苗在理想、干旱和自然三种生长条件下的图像数据集，共计7,590张图像，标注了种类标签和边界框，适用于植物分类和检测任务。