JMuBEN

The aim of creating JMuBEN dataset is to add to the limited availability of Arabica leaf images that are available online. This dataset will help researchers to be able to test the accuracy of their deep learning models without spending much time going to the field for data collection. The Arabica dataset (JMuBEN) contains images that are useful in training and validation during the utilization of deep learning algorithms used in plant disease recognition and classification. The dataset contains leaf images which were collected from Arabica coffee type and it shows three sets of unhealthy images. In total, 22591 images of Arabica coffee are included in JMuBEN dataset.
The data has been cropped to emphasize the region of interest where the disease is seen to reduce the training time by avoiding background learning. The images were also resized to maintain uniformity in shape and size. The images that were smaller were augmented with the aim of preventing over-fitting during training and validation of models. There are a total of 3 files having images of coffee leaves affected by Coffee Rust, Cescospora and Phoma .

创建JMuBEN数据集的宗旨在于补充网络中稀少的阿拉比卡咖啡叶图像资源。该数据集旨在协助研究人员在无需花费大量时间前往实地收集数据的情况下，对深度学习模型的准确性进行测试。Arabica数据集（JMuBEN）包含了在深度学习算法应用于植物病害识别与分类过程中，用于训练与验证的有用图像。该数据集收纳了从阿拉比卡咖啡类型中收集的叶片图像，并展示了三组不健康图像。JMuBEN数据集中总计包含了22591张阿拉比卡咖啡叶片图像。数据经过裁剪，以突出显示病害出现的兴趣区域，从而通过避免背景学习来减少训练时间。图像亦经过调整尺寸，以保持形状和尺寸的统一性。对于尺寸较小的图像，通过数据增强技术，旨在防止模型在训练与验证过程中的过拟合。总计有3个文件，包含了受咖啡锈病、Cescospora和Phoma影响的咖啡叶片图像。