Scholarlink

<p>Citrus foliar diseases such as Black Spot, Canker, Greening, Scab, and Melanose significantly impact fruit yield and quality, underscoring the need for accurate and automated detection techniques. This study presents a hybrid deep learning framework that integrates EfficientNetB0 and MobileNetV2 in a dual-stream architecture for the classification of citrus leaf diseases. The proposed methodology comprises image acquisition from publicly available datasets, preprocessing through augmentation and enhancement, segmentation of leaf regions, and classification using the hybrid model. The novelty of this work lies in the dual-stream integration of EfficientNetB0 and MobileNetV2, which combines the superior feature extraction capacity of EfficientNetB0 with the lightweight computational efficiency of MobileNetV2. This fusion enables richer feature representation and faster convergence compared to conventional single-network or handcrafted feature-based approaches. Experimental evaluation reveals that MobileNetV2 achieved an accuracy of 73.77%, while EfficientNetB0 improved performance to 92.62%. The hybrid architecture further enhanced accuracy to 94.26%, with precision, recall, and F1-score exceeding 96%. Incorporating augmented data elevated the model’s performance to 99.80% accuracy, demonstrating strong generalisation capability. In comparison with existing CNN-based models such as VGG16, ResNet50, and InceptionV3, the proposed hybrid framework exhibits superior accuracy, reduced computational complexity, and enhanced robustness. Overall, the hybrid model provides a practical and scalable solution for early detection of citrus diseases, contributing to sustainable crop protection and improved agricultural productivity.</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

Enhancing citrus crop protection via hybrid CNN model combining EfficientNetB0 and MobileNetV2

  • 期刊论文

  • 发表日期：

    2025年12月20日

查看全文
  • Anuj Kumar,
  • Rishab Batra

摘要

Citrus foliar diseases such as Black Spot, Canker, Greening, Scab, and Melanose significantly impact fruit yield and quality, underscoring the need for accurate and automated detection techniques. This study presents a hybrid deep learning framework that integrates EfficientNetB0 and MobileNetV2 in a dual-stream architecture for the classification of citrus leaf diseases. The proposed methodology comprises image acquisition from publicly available datasets, preprocessing through augmentation and enhancement, segmentation of leaf regions, and classification using the hybrid model. The novelty of this work lies in the dual-stream integration of EfficientNetB0 and MobileNetV2, which combines the superior feature extraction capacity of EfficientNetB0 with the lightweight computational efficiency of MobileNetV2. This fusion enables richer feature representation and faster convergence compared to conventional single-network or handcrafted feature-based approaches. Experimental evaluation reveals that MobileNetV2 achieved an accuracy of 73.77%, while EfficientNetB0 improved performance to 92.62%. The hybrid architecture further enhanced accuracy to 94.26%, with precision, recall, and F1-score exceeding 96%. Incorporating augmented data elevated the model’s performance to 99.80% accuracy, demonstrating strong generalisation capability. In comparison with existing CNN-based models such as VGG16, ResNet50, and InceptionV3, the proposed hybrid framework exhibits superior accuracy, reduced computational complexity, and enhanced robustness. Overall, the hybrid model provides a practical and scalable solution for early detection of citrus diseases, contributing to sustainable crop protection and improved agricultural productivity.