Scholarlink

<p>Nematodes are microscopic organisms that play dual roles in agriculture: while some act as natural biocontrol agents, others-particularly Plant-Parasitic Nematode (PPN)-cause substantial crop damage, leading to an estimated 10%–20% reduction in global agricultural yields annually. Accurate species-level identification is thus critical for both leveraging beneficial nematodes and mitigating harmful ones. Traditional identification methods, however, are time-consuming, labor-intensive, and reliant on expert knowledge and laboratory infrastructure. This paper uses weighted average ensemble techniques to provide an automated classification framework for PPNs based on deep learning in order to get over these limitations. The framework employs a weighted average fusion strategy to integrate predictions from three pre-trained Convolutional Neural Networks (CNNs), namely MobileNetV2, ResNet101, and InceptionV3, thereby enhancing classification performance. There are 17 distinct PPN species included in the 2,500 microscopic images in the dataset. Experimental evaluation on the test set demonstrates that the ensemble model achieves an average classification accuracy of 98%, outperforming individual models (MobileNetV2: 95%, InceptionV3: 92%, ResNet101: 95%). These results highlight the effectiveness of ensemble learning combined with transfer learning in delivering accurate, scalable, and practical solutions for nematode identification in agricultural diagnostics.</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

Deep transfer learning-based ensemble framework for identifying plant-parasitic nematodes from microscopic images

  • 期刊论文

  • 发表日期：

    2026年3月9日

查看全文
  • Meetali Verma,
  • Rahul Sharma,
  • Ayushi Kotwal,
  • Jatinder Manhas,
  • Vinod Sharma

摘要

Nematodes are microscopic organisms that play dual roles in agriculture: while some act as natural biocontrol agents, others-particularly Plant-Parasitic Nematode (PPN)-cause substantial crop damage, leading to an estimated 10%–20% reduction in global agricultural yields annually. Accurate species-level identification is thus critical for both leveraging beneficial nematodes and mitigating harmful ones. Traditional identification methods, however, are time-consuming, labor-intensive, and reliant on expert knowledge and laboratory infrastructure. This paper uses weighted average ensemble techniques to provide an automated classification framework for PPNs based on deep learning in order to get over these limitations. The framework employs a weighted average fusion strategy to integrate predictions from three pre-trained Convolutional Neural Networks (CNNs), namely MobileNetV2, ResNet101, and InceptionV3, thereby enhancing classification performance. There are 17 distinct PPN species included in the 2,500 microscopic images in the dataset. Experimental evaluation on the test set demonstrates that the ensemble model achieves an average classification accuracy of 98%, outperforming individual models (MobileNetV2: 95%, InceptionV3: 92%, ResNet101: 95%). These results highlight the effectiveness of ensemble learning combined with transfer learning in delivering accurate, scalable, and practical solutions for nematode identification in agricultural diagnostics.