Scholarlink

The agriculture sector in sub-Saharan Africa faces significant challenges in growing vital food security crops like maize due to crop diseases. In response, we introduce a robust deep learning model for accurately identifying diseased and healthy maize leaves using leaf images. Specifically, our model distinguishes between maize leaves affected by Maize Lethal Necrosis (MLN), Maize Streak Virus (MLV) and healthy ones. We utilize the ResNet architecture, a well-established convolutional neural network known for its exceptional performance in computer vision tasks. Our experiments demonstrate that using a pretrained ResNet model achieves remarkable accuracy, exceeding 99%, in distinguishing between diseased and healthy maize leaves. Importantly, we find that employing a pre-trained Deep Learning (DL) model eliminates the need for extensive training from scratch, affirming the effectiveness of Transfer Learning (TL) in this field. Furthermore, our model proves its robustness by effectively handling diverse variations in leaf appearance, lighting conditions, and disease symptoms. This research offers practical tools for early disease detection, aiding farmers, and agricultural experts, ultimately enhancing maize production. Our results highlight the potential of harnessing DL techniques to address critical challenges in agriculture and food security.

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 Maize Crop Health: Deep Learning Approach for Disease Detection and Classification Using Leaf Images

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Roseline Nyange,
  • Mapopa Gota Chipofya,
  • Srishti Goel,
  • S. B. Ashoka,
  • Channabasava Chola

摘要

The agriculture sector in sub-Saharan Africa faces significant challenges in growing vital food security crops like maize due to crop diseases. In response, we introduce a robust deep learning model for accurately identifying diseased and healthy maize leaves using leaf images. Specifically, our model distinguishes between maize leaves affected by Maize Lethal Necrosis (MLN), Maize Streak Virus (MLV) and healthy ones. We utilize the ResNet architecture, a well-established convolutional neural network known for its exceptional performance in computer vision tasks. Our experiments demonstrate that using a pretrained ResNet model achieves remarkable accuracy, exceeding 99%, in distinguishing between diseased and healthy maize leaves. Importantly, we find that employing a pre-trained Deep Learning (DL) model eliminates the need for extensive training from scratch, affirming the effectiveness of Transfer Learning (TL) in this field. Furthermore, our model proves its robustness by effectively handling diverse variations in leaf appearance, lighting conditions, and disease symptoms. This research offers practical tools for early disease detection, aiding farmers, and agricultural experts, ultimately enhancing maize production. Our results highlight the potential of harnessing DL techniques to address critical challenges in agriculture and food security.