Classification and Detection of Poultry Disease from Chicken Fecal Images Using Deep Learning Techniques

Poultry diseases significantly threaten global food security and the poultry industry, often resulting in economic losses and elevated mortality rates. Early and accurate disease detection is essential to mitigating outbreaks, yet traditional diagnostic methods remain time-intensive and error-prone. This study leverages deep learning techniques to classify and detect poultry diseases from chicken fecal images, aiming to provide a scalable and automated solution. A dataset comprising four classes Healthy, Coccidiosis, Salmonella, and Newcastle Disease was utilised for training and evaluation. DenseNet-121 with Squeeze-and-Excitation (SE) blocks demonstrated outstanding performance in classification, achieving 97.75% accuracy. For disease localisation, YOLOv8 was employed, yielding high Mean Average Precision (mAP) and Precision-Recall metrics. These models were integrated into Streamlit and Flask applications for real-time deployment, offering an efficient and accessible solution for poultry health monitoring. This work highlights the feasibility of AI-powered disease detection, paving the way for advanced field applications and improved disease management strategies.