Strawberry growth state dataset

The dataset used in this study is sourced from two public strawberry datasets: StrawDI_Db1 and the Strawberry Dataset for Object Detection. StrawDI_Db1 consists of 3,803 images captured in strawberry plantations in Huelva, Spain, during different harvesting activities. These images have a resolution of 4032×3024 pixels and are stored in 8-bit per channel JPEG format. Upon release, they were resized to 1008×756 pixels and saved in PNG format. The Strawberry Dataset for Object Detection , released by the Natural Resources Institute Finland (Luke) in 2022, contains 813 images for strawberry flower and fruit object detection.After obtaining the datasets, we performed image cleaning, removing blurry and unidentifiable images, and merged them to create a final set of 2,301 images covering multiple growth stages of strawberry flowers and fruits. The merged dataset was split into training, and test sets at a ratio of 8:2, ensuring no overlapping images between subsets, and the test set was not involved in training to ensure the rigor of model evaluation.

本研究使用的数据集源自两个公开草莓数据集：StrawDI_Db1，以及面向目标检测的草莓数据集（Strawberry Dataset for Object Detection）。StrawDI_Db1包含3803张采集自西班牙韦尔瓦（Huelva）草莓种植园的图像，拍摄场景涵盖不同采收作业时段。这批图像的分辨率为4032×3024像素，采用每通道8位的JPEG格式存储；该数据集发布时，图像已被调整至1008×756像素，并以PNG格式存储。面向目标检测的草莓数据集（Strawberry Dataset for Object Detection）由芬兰自然资源研究所（Natural Resources Institute Finland，Luke）于2022年发布，包含813张用于草莓花与果实目标检测的图像。在获取原始数据集后，我们开展了图像清洗工作，剔除模糊不清、无法识别的图像，并将剩余数据合并，最终得到涵盖草莓花与果实多个生长阶段的2301张图像组成的最终数据集。我们将合并后的数据集以8:2的比例划分为训练集与测试集，确保各子集间无图像重叠，且测试集未参与模型训练，以保障模型评估的严谨性。