Scholarlink

Poultry, particularly chicken, serves as a primary source of animal protein in our daily lives, but it faces persistent threats from many poultry diseases. Poultry farmers face challenges from poultry diseases, which often lead to the widespread outbreak of poultry epidemics. Fortunately, timely detection of poultry diseases is possible for early diagnosis and treatment to prevent big financial losses. Therefore, Artificial Intelligence (AI) in healthcare provides a promising solution for timely poultry disease detection and prevention for diseases like Coccidiosis, Newcastle Disease, Salmonella among others. This chapter introduces disease detection models i.e., MobileNet and a CNN with 97.00 and 92% precision based on image classification of poultry fecal matter seeking to revolutionize current practices that enabled even non-specialist farmers to proactively detect and diagnose poultry diseases using Mobile phones. This innovative approach not only empowers farmers digital tools to safeguard their flocks efficiently but also addresses the black box nature of AI models by enhancing model trustworthiness using class activation maps to provide explanations for the reason behind each diagnosis. The visual explanations provided by class activation maps ensured that farmers, including those without a scientific background, gain insights into the factors that influenced disease diagnostics. This transparency enhances the comprehensibility and trustworthiness of the model outcomes thus signifying a critical step toward responsible intelligent software systems that inclusively empower farmers to develop clear judgments based on an explicit comprehension of the model’s reasoning processes.

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

Trustworthy Intelligent Poultry Disease Diagnostics Using Class Activation Maps For Deep Learning Visual Explanations

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Bukenya Abdusalaam,
  • Bogera Henry,
  • Binomugisha Bariyo Brian,
  • Ssekyanzi Joel,
  • Rose Nakibuule,
  • Nakayiza Hellen,
  • Ggaliwango Marvin

摘要

Poultry, particularly chicken, serves as a primary source of animal protein in our daily lives, but it faces persistent threats from many poultry diseases. Poultry farmers face challenges from poultry diseases, which often lead to the widespread outbreak of poultry epidemics. Fortunately, timely detection of poultry diseases is possible for early diagnosis and treatment to prevent big financial losses. Therefore, Artificial Intelligence (AI) in healthcare provides a promising solution for timely poultry disease detection and prevention for diseases like Coccidiosis, Newcastle Disease, Salmonella among others. This chapter introduces disease detection models i.e., MobileNet and a CNN with 97.00 and 92% precision based on image classification of poultry fecal matter seeking to revolutionize current practices that enabled even non-specialist farmers to proactively detect and diagnose poultry diseases using Mobile phones. This innovative approach not only empowers farmers digital tools to safeguard their flocks efficiently but also addresses the black box nature of AI models by enhancing model trustworthiness using class activation maps to provide explanations for the reason behind each diagnosis. The visual explanations provided by class activation maps ensured that farmers, including those without a scientific background, gain insights into the factors that influenced disease diagnostics. This transparency enhances the comprehensibility and trustworthiness of the model outcomes thus signifying a critical step toward responsible intelligent software systems that inclusively empower farmers to develop clear judgments based on an explicit comprehension of the model’s reasoning processes.