马铃薯缺陷目标检测

本研究所使用的图像样本主要由收录于Mendeley Data的马铃薯缺陷数据集混合而成，具有良好的标注质量，适用于马铃薯缺陷目标检测任务。主要包括以下两个部分：由Md Ali Emam Al等于2024年发布的Potato Disease Classification数据集，以及由Md Mafiul Hasan Matin Mafi等于2023年发布的Potato Disease Recognition Dataset。
研究团队在此基础上收集并整合了马铃薯原始缺陷图像共计2328张。其包含健康马铃薯（Good）、机械损伤马铃薯（Defected potato）、真菌感染马铃薯（Diseased-fungal-damaged）与发芽马铃薯（Bud-Sprouted）四种样本类别。之后对图像进行统一裁剪与尺寸标准化，以保证输入一致性，为后续标注与模型训练奠定数据基础。

The image samples used in this study are mainly compiled from potato defect datasets deposited in Mendeley Data, which feature high-quality annotations and are suitable for potato defect object detection tasks. These datasets mainly include two parts: the Potato Disease Classification dataset released by Md Ali Emam Al et al. in 2024, and the Potato Disease Recognition Dataset released by Md Mafiul Hasan Matin Mafi et al. in 2023.
The research team collected and integrated a total of 2328 original potato defect images based on these datasets. These images cover four sample categories: Good (healthy potatoes), Defected potato (mechanically damaged potatoes), Diseased-fungal-damaged (potatoes damaged by fungal infection), and Bud-Sprouted (sprouted potatoes). Subsequently, uniform cropping and size standardization were conducted on all images to ensure consistent input, laying a solid data foundation for subsequent annotation and model training.

搜集汇总

背景与挑战

背景概述

该数据集专注于马铃薯缺陷的目标检测任务，涉及图像识别和网络预训练技术，但当前在平台中无具体数据、介绍或下载资源可供访问。