Scholarlink

<p>This paper presents CropLeafNet, a deep learning framework for accurate and scalable plant disease detection in real-world agricultural conditions. The system integrates publicly available datasets of Plant Village PlantVillage Dataset (PlantVillage, Pennsylvania State University, 2025) with real-time field images of cotton and castor leaves collected across Gujarat, improving robustness to variations in lighting, background, and crop diversity. A novel Sliding Window Mean Average Deviation (SWMAD) preprocessing algorithm enhances brightness and color correction, highlighting diseased regions. A hybrid convolutional neural network (CNN) segmentation model with convolution–deconvolution layers achieves precise, noise-resilient segmentation. The dual-layer classification architecture, with independent CNNs for plant type and disease recognition, reduces inter-class confusion and enables multi-crop scalability. CropLeafNet is deployed as a lightweight TensorFlow Lite (TFLite) Android application for real-time, offline detection. Experimental results demonstrate 98% classification accuracy, confirming robustness and practical applicability.</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

A dynamic deep learning framework for real-time multi-plant, multi-disease detection under diverse environmental conditions

  • 期刊论文

  • 发表日期：

    2025年12月20日

查看全文
  • Sejal Rahul Trivedi,
  • Neha Sharma

摘要

This paper presents CropLeafNet, a deep learning framework for accurate and scalable plant disease detection in real-world agricultural conditions. The system integrates publicly available datasets of Plant Village PlantVillage Dataset (PlantVillage, Pennsylvania State University, 2025) with real-time field images of cotton and castor leaves collected across Gujarat, improving robustness to variations in lighting, background, and crop diversity. A novel Sliding Window Mean Average Deviation (SWMAD) preprocessing algorithm enhances brightness and color correction, highlighting diseased regions. A hybrid convolutional neural network (CNN) segmentation model with convolution–deconvolution layers achieves precise, noise-resilient segmentation. The dual-layer classification architecture, with independent CNNs for plant type and disease recognition, reduces inter-class confusion and enables multi-crop scalability. CropLeafNet is deployed as a lightweight TensorFlow Lite (TFLite) Android application for real-time, offline detection. Experimental results demonstrate 98% classification accuracy, confirming robustness and practical applicability.