LIGHTWEIGHT DEEP LEARNING BASED AUTOMATED POULTRY DISEASE DIAGNOSIS SYSTEM

Abstract

Poultry farming plays a vital role in ensuring global food security; however, infectious diseases remain a persistent threat to flock health, productivity, and economic stability. Conventional poultry disease diagnosis depends on visual assessment by farm workers and laboratory-based tests such as bacterial cultures and PCR, which are often expensive, time-consuming, and inaccessible to farmers in rural or resource-limited regions. These delays in diagnosis allow diseases such as coccidiosis, Newcastle disease, and avian influenza to spread rapidly, increasing mortality rates and antibiotic usage. Moreover, early-stage symptoms are often subtle and easily missed, reducing the reliability of manual screening. To address these challenges, this research proposes an automated, image-based poultry disease diagnostic system using deep learning and transfer learning techniques. The system integrates pre-trained convolutional neural network models InceptionV3, MobileNetV2, and VGG16 to efficiently extract disease-specific features from poultry images while maintaining manageable computational requirements. A Tkinter-based graphical user interface enables farmers to upload images, perform analysis, and receive rapid diagnostic feedback without specialized technical knowledge. By providing on-farm, real-time disease classification, the proposed system reduces dependence on veterinary laboratories, shortens diagnostic turnaround time, and supports timely disease control decisions. Overall, this work contributes a practical, accessible, and accurate solution for improving poultry health management and minimizing economic losses.