BD Guava Fruit and Leaf Disease Dataset

Bangladesh, being a major guava-producing country, relies heavily on seasonal harvests.
However, diseases such as algal leaf spot, red rust, scab, and black mold severely impact guava
yield and farmer income [4]. Early detection is essential to prevent large-scale infection and
ensure sustainable fruit production. With the increasing availability of affordable computing
power and image-based data, deep learning provides a powerful solution for rapid and accurate
disease identification.
Previous works have demonstrated the high potential of deep learning models in similar
agricultural domains. For example, YOLOv8 has been successfully applied in apple orchard
segmentation tasks, outperforming Mask R-CNN in both accuracy and inference speed,
achieving precision up to 0.93 and recall 0.97, proving its robustness for real-time agricultural
applications [5]. Similarly, in tomato grading, YOLOv7 achieved 99.2% accuracy and 99.4%
recovery rate in detecting and classifying tomato quality, highlighting the power of convolutional
deep learning in fruit analysis [6].
Inspired by these advancements, this study applies YOLO and CNN models for guava leaf and
fruit disease detection to enhance precision agriculture in Bangladesh. The integration of these
models into user-friendly platforms such as Tkinter and Gradio aims to assist farmers and
researchers in diagnosing diseases efficiently, thereby reducing crop losses and improving yield
quality.
Version 2 Red G
The dataset is organized into multiple class-wise folders, with images standardized to a uniform resolution for consistency. Each class contains a sufficient number of samples to support effective model training and evaluation. The data is structured into separate subsets for training, validation, and testing, while preserving class-wise organization to ensure balanced and reliable performance assessment.
Version 3 all including
The dataset was expanded to increase sample diversity and improve model generalization.
Image variations were generated through multiple controlled transformations applied to the original data.
The final dataset consists of ten distinct categories with a balanced number of samples per category.
The data is organized in a structured format suitable for training, validation, and testing in deep learning models.
CODE: https://github.com/shuvobasak4004/Guava-Multi-Model-Training-Code

孟加拉国作为番石榴生产大国，高度依赖季节性收获。然而，藻斑病、红锈病、疮痂病和黑霉病等病害严重影响番石榴产量与农户收益[4]。早期检测对于防止大规模感染、保障果品可持续生产至关重要。随着低成本算力与图像数据的可及性不断提升，深度学习为快速精准的病害识别提供了强有力的解决方案。

已有研究证明深度学习模型在同类农业领域具备极高应用潜力。例如，YOLOv8已成功应用于苹果园分割任务，在精度与推理速度上均优于Mask R-CNN，其精确率可达0.93、召回率达0.97，证实了其在实时农业应用中的鲁棒性[5]。类似地，在番茄分级任务中，YOLOv7在番茄品质检测与分类任务中实现了99.2%的准确率与99.4%的召回率，凸显了卷积神经网络（Convolutional Neural Network, CNN）在果品分析中的应用价值[6]。

受这些研究进展启发，本研究将YOLO与卷积神经网络模型应用于番石榴叶片与果实病害检测，以助力孟加拉国的精准农业发展。本研究将这些模型集成至Tkinter、Gradio等易用平台，旨在帮助农户与科研人员高效完成病害诊断，从而减少作物损失、提升果品质量。

版本2 红G
本数据集按类别划分文件夹，图像统一分辨率以保证一致性。每个类别包含充足样本以支撑模型的有效训练与评估。数据被划分为训练集、验证集与测试集三个子集，同时保留按类别组织的结构，以确保性能评估的平衡性与可靠性。

版本3 全类别集成
本数据集通过扩充样本多样性以提升模型泛化能力。通过对原始数据施加多种可控变换生成图像变体。最终数据集包含10个不同类别，且各类别样本量保持均衡。该数据以适配深度学习模型训练、验证与测试的结构化格式进行组织。

代码仓库：https://github.com/shuvobasak4004/Guava-Multi-Model-Training-Code