Scholarlink

Food production is essential for human survival, not only in India but across the globe. Both vegetables and fruits serve as vital sources of nutrients and minerals. Among fruits, mangoes hold significant importance as they provide essential vitamins (A, C, E, and B6) that help the body combat various diseases and strengthen immunity. However, mango foliage pathologies threaten agricultural sustainability and human health, impacting economic stability worldwide. Ineffective disease management among mango crops can lead to substantial financial losses, reduced crop yields, and increased management costs. As a major crop in many countries, mangoes play a crucial role in supporting local livelihoods, facilitating international trade, and enhancing food security. In this research, we propose an approach for identifying and detecting mango leaf disorders using several CNN-based pre-trained models, including EfficientNetB0, DenseNet121, DenseNet201, MobileNetV2, ResNet50, ResNet101, VGG16, and VGG19. These models incorporate a comprehensive set of operations, such as data acquisition, preprocessing, and the division of the data repository into training, validation, and testing subsets, along with data augmentation to enhance performance. After performing these steps, EfficientNetB0 achieved impressive metrics: 99.72% accuracy, 99.72% precision, 100% recall, 99.72% F1 score, and a 100% AUC-ROC curve, signifying its superior performance compared to the other pre-trained models.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Optimizing Mango Leaf Disease Detection and Diagnosis Through Data Augmentation and Pre-trained CNN Models

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Hardeep Kaur,
  • Komalpreet Kaur

摘要

Food production is essential for human survival, not only in India but across the globe. Both vegetables and fruits serve as vital sources of nutrients and minerals. Among fruits, mangoes hold significant importance as they provide essential vitamins (A, C, E, and B6) that help the body combat various diseases and strengthen immunity. However, mango foliage pathologies threaten agricultural sustainability and human health, impacting economic stability worldwide. Ineffective disease management among mango crops can lead to substantial financial losses, reduced crop yields, and increased management costs. As a major crop in many countries, mangoes play a crucial role in supporting local livelihoods, facilitating international trade, and enhancing food security. In this research, we propose an approach for identifying and detecting mango leaf disorders using several CNN-based pre-trained models, including EfficientNetB0, DenseNet121, DenseNet201, MobileNetV2, ResNet50, ResNet101, VGG16, and VGG19. These models incorporate a comprehensive set of operations, such as data acquisition, preprocessing, and the division of the data repository into training, validation, and testing subsets, along with data augmentation to enhance performance. After performing these steps, EfficientNetB0 achieved impressive metrics: 99.72% accuracy, 99.72% precision, 100% recall, 99.72% F1 score, and a 100% AUC-ROC curve, signifying its superior performance compared to the other pre-trained models.