Scholarlink

<p>Mango (<i>Mangifera indica</i>&#xa0;L.), a&#xa0;tropical fruit of significant agricultural and economic value, is greatly affected by leaf diseases such as anthracnose, bacterial canker, cutting weevil, dieback, gall midge, powdery mildew, and sooty mold, which reduce crop yield and fruit quality. This work presents an ensemble ResVGG-Net model merging fine-tuned ResNet50 and VGG16 architectures for the categorization of mango leaf diseases into eight categories, including healthy leaves, in order to overcome the limits of conventional manual inspection approaches. Using their residual connections for strong high-level feature abstraction, the ResNet50 and VGG16 architectures were improved by including GlobalAveragePooling2D, batch normalization, dropout layers, and dense layers with ReLU activation. Concatenation merged the characteristics taken from both fine-tuned models to produce a&#xa0;single representation using the complementary strengths of both architectures. Furthermore, to improve the concatenated features and guarantee accurate classification, bespoke layers were used comprising completely linked dense layers with dropout and batch normalization. Including a&#xa0;4000-image balanced dataset with 500 images per class, the model surpassed individual models in robustness and performance by achieving an accuracy of 99.88% and an overall F1&#xa0;score of 0.9975, confirming reliable classification across all disease categories. Early disease detection is achieved through the proposed ResVGG-Net model as a&#xa0;scalable and autonomous solution that promotes sustainable agricultural practices by enabling timely interventions and reducing crop losses. Furthermore, the method is applicable in agricultural diagnostics and precision farming and may be expanded to other crops.</p>

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

Proposed ResVGG-Net Model for Mango Leaf Disease Classification and Agricultural Sustainability

  • 期刊论文

  • 发表日期：

    2025年11月6日

查看全文
  • Vishnu Kant,
  • Sheifali Gupta,
  • Deepali Gupta,
  • Sapna Juneja,
  • Yonis Gulzar,
  • Niranjan Devanathan R.

摘要

Mango (Mangifera indica L.), a tropical fruit of significant agricultural and economic value, is greatly affected by leaf diseases such as anthracnose, bacterial canker, cutting weevil, dieback, gall midge, powdery mildew, and sooty mold, which reduce crop yield and fruit quality. This work presents an ensemble ResVGG-Net model merging fine-tuned ResNet50 and VGG16 architectures for the categorization of mango leaf diseases into eight categories, including healthy leaves, in order to overcome the limits of conventional manual inspection approaches. Using their residual connections for strong high-level feature abstraction, the ResNet50 and VGG16 architectures were improved by including GlobalAveragePooling2D, batch normalization, dropout layers, and dense layers with ReLU activation. Concatenation merged the characteristics taken from both fine-tuned models to produce a single representation using the complementary strengths of both architectures. Furthermore, to improve the concatenated features and guarantee accurate classification, bespoke layers were used comprising completely linked dense layers with dropout and batch normalization. Including a 4000-image balanced dataset with 500 images per class, the model surpassed individual models in robustness and performance by achieving an accuracy of 99.88% and an overall F1 score of 0.9975, confirming reliable classification across all disease categories. Early disease detection is achieved through the proposed ResVGG-Net model as a scalable and autonomous solution that promotes sustainable agricultural practices by enabling timely interventions and reducing crop losses. Furthermore, the method is applicable in agricultural diagnostics and precision farming and may be expanded to other crops.