The Growing Strawberries Dataset

The Growing Strawberries Dataset (GSD) is a curated multiple-object tracking dataset inspired by the growth monitoring of strawberries. The frames were taken at hourly intervals by six cameras for in total of 16 months in 2021 and 2022, covering 12 plants in two greenhouses respectively. The dataset consists of hourly images collected during the cultivation period, bounding box (bbox) annotations of strawberry fruits, and precise identification and tracking of strawberries over time. GSD contains two types of images - RGB (visual spectrum) and OCN (orange, cyan, near-infrared). These images were captured throughout the cultivation period. Each image sequence represents all the images captured by one camera during the year of cultivation. These sequences are named using the format "<RGB/OCN-1/2/3-2021/2022>." In addition, a small sample captured by one of the RGB cameras is included, featuring images and annotations from a two-day period, along with a demo video spanning seven days. The ground-truth annotations of the bbox and trajectories of strawberries are provided with a coco-format JSON file and a TXT file for compatibility with the MOT Challenge evaluation tools. In addition, we archived further (non-annotated) growth monitoring from 2023 and 2024 and images that were too dark in other datasets of the same collection.

草莓生长数据集（Growing Strawberries Dataset，缩写GSD）是一款专为草莓生长监测任务打造的多目标跟踪数据集。该数据集的影像由6台摄像头以每小时1帧的频率采集，采集周期覆盖2021年至2022年共计16个月，分别涵盖两个温室中的12株草莓植株。数据集包含栽培周期内每小时采集的影像、草莓果实的边界框（bounding box，简称bbox）标注，以及草莓在全周期内的精准识别与跟踪结果。GSD包含两类影像：RGB（可见光光谱）与OCN（橙、青、近红外多光谱），所有影像均采集自草莓栽培全周期。每一个影像序列对应单台摄像头在单一年度栽培周期内采集的全部影像，此类序列的命名格式为"<RGB/OCN-1/2/3-2021/2022>"。此外，数据集还包含一台RGB摄像头采集的小型样本集，涵盖为期两天的影像与对应标注，同时附带一段时长七天的演示视频。草莓边界框与运动轨迹的真值标注采用两种格式提供：COCO格式的JSON文件，以及适配多目标跟踪挑战赛（MOT Challenge）评估工具的TXT文件。此外，本项目还归档了2023年至2024年的非标注草莓生长监测数据，以及同批次采集数据集内过暗的影像。

搜集汇总

背景与挑战

背景概述

The Growing Strawberries Dataset是一个用于多目标跟踪的草莓生长监测数据集，包含2021年和2022年16个月内每小时采集的RGB和OCN图像，覆盖12株植物，并提供边界框和轨迹注释。该数据集专为计算机视觉和园艺研究设计，支持长期生物发育跟踪，适用于人工智能和农业科学领域。