MangoLeafDB-C
收藏资源简介:
MangoLeafDB-C是由阿拉戈斯联邦大学团队基于原始MangoLeafDB数据集构建的芒果叶病害分析增强数据集,包含8类病害的4000张叶片图像。该数据集通过引入19种人工合成 corruption(如模糊、噪声、光照变化等)在5种严重程度下的组合,扩展为95个子集,旨在模拟真实农业场景中的图像退化情况。数据来源于孟加拉国芒果种植区的实地采集,经标准化预处理至224×224像素分辨率。该数据集主要用于评估卷积神经网络在农业边缘计算场景下对叶片病害诊断的鲁棒性,为解决实际应用中因图像质量波动导致的模型性能下降问题提供基准测试平台。
MangoLeafDB-C is an enhanced dataset for mango leaf disease analysis developed by the team from Federal University of Alagoas based on the original MangoLeafDB dataset. It contains 4000 leaf images across 8 disease categories. This dataset is expanded into 95 subsets by incorporating 19 types of artificially synthesized image corruptions (including blur, noise, illumination variation, etc.) across 5 severity levels, with the goal of simulating image degradation conditions in real-world agricultural scenarios. The data was field-collected from mango planting areas in Bangladesh, and standardized and preprocessed to a resolution of 224×224 pixels. This dataset is primarily used to evaluate the robustness of convolutional neural networks (CNNs) for leaf disease diagnosis in agricultural edge computing scenarios, serving as a benchmark platform to address the problem of model performance degradation caused by fluctuating image quality in practical applications.




