Chicktech AI: Automated Detection of Poultry Diseases Through Transfer Learning and Multi-Class Convolutional Neural Network Classification

Abstract

Poultry production is a critical component of global food security, yet disease outbreaks remain a persistent challenge for smallholder farmers who lack timely and affordable diagnostic tools. Diseases such as coccidiosis, bumblefoot, and fowlpox cause severe financial losses and high flock mortality. This paper presents ChickTech AI, a fully deployed, cloud-based poultry disease-diagnostics platform that integrates deep learning with a multilingual, mobile-responsive web interface. The system employs a dual-pipeline transfer learning architecture: VGG16 for fecal texture analysis (coccidiosis detection) and MobileNetV2 for external lesion classification (bumblefoot and fowlpox), both fine-tuned using a progressive three-phase training strategy. Evaluated on a held-out test set of 465 images, the system achieves an overall classification accuracy of 98.2% with all-class AUC > 0.96. The platform incorporates multilingual voice interaction (speech-to-text, text-to-speech, and batch translation) and a tiered clinical decision support module (CRITICAL / IMPORTANT / SUPPORTIVE) that translates AI predictions into actionable treatment guidance. Unlike prior works that report only algorithmic accuracy, ChickTech AI bridges the research-to-practice gap through a production-ready, cloud-hosted deployment accessible via smartphone, demonstrating that AI-driven precision livestock diagnostics can be scalable, inclusive, and practically deployable.