Scholarlink

Poultry farming is a vital sector for global food security, yet it is highly susceptible to diseases that can cause significant economic losses. Traditional disease detection methods, such as manual inspection and laboratory analysis, are time-consuming and prone to errors. This paper proposes an advanced deep learning-based approach for poultry disease detection using Convolutional Neural Networks (CNNs) and a series of image preprocessing techniques. The proposed pipeline involves background removal, emboss filter application, and histogram equalization to enhance critical visual features for improved disease detection accuracy. Various deep learning models, including ResNet 152, MobileNet V2, VGG16, EfficientNet B7, and EfficientNet B0, were evaluated, with EfficientNet B0 achieving the highest accuracy of 98.0%. Additionally, the average ensemble approach, combining EfficientNet B7 and EfficientNet B0, achieved an accuracy of 98.46%, demonstrating the effectiveness of model ensembling. The results highlight the potential of AI-driven solutions to revolutionize disease detection in poultry farming, providing faster, more accurate alternatives to traditional methods and contributing to improved disease management 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

Optimized Deep Learning Approach for Accurate Poultry Disease Detection

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Imranul Islam Adnan,
  • Mohammed Z. Waughfa,
  • Ehfaz Faisal Mahee,
  • Saieef Sunny,
  • Md. Mehedi Kayser Chowdhury

摘要

Poultry farming is a vital sector for global food security, yet it is highly susceptible to diseases that can cause significant economic losses. Traditional disease detection methods, such as manual inspection and laboratory analysis, are time-consuming and prone to errors. This paper proposes an advanced deep learning-based approach for poultry disease detection using Convolutional Neural Networks (CNNs) and a series of image preprocessing techniques. The proposed pipeline involves background removal, emboss filter application, and histogram equalization to enhance critical visual features for improved disease detection accuracy. Various deep learning models, including ResNet 152, MobileNet V2, VGG16, EfficientNet B7, and EfficientNet B0, were evaluated, with EfficientNet B0 achieving the highest accuracy of 98.0%. Additionally, the average ensemble approach, combining EfficientNet B7 and EfficientNet B0, achieved an accuracy of 98.46%, demonstrating the effectiveness of model ensembling. The results highlight the potential of AI-driven solutions to revolutionize disease detection in poultry farming, providing faster, more accurate alternatives to traditional methods and contributing to improved disease management and food security.