AI-DISA: An Artificial Intelligence-Based Disease Identification System for Livestock Health Management
摘要
Livestock health plays a pivotal role in ensuring global food security, sustainable agriculture, and rural livelihoods. The increasing need for timely and accessible veterinary diagnostics has catalyzed the integration of Artificial Intelligence (AI) into livestock disease management. This chapter introduces AI-DISA (Artificial Intelligence-based Disease Identification System for Animals), India’s first mobile-based AI solution for livestock disease detection, developed by ICAR-IASRI in collaboration with ICAR-IVRI. AI-DISA leverages deep learning models—particularly Convolutional Neural Networks (CNNs)—trained on a curated dataset from the National Image Base for Livestock Diseases (NIBLD). It incorporates both object detection and image classification approaches using advanced architectures such as SSD, YOLO, and RetinaNet, enabling precise localization and identification of disease symptoms from livestock images. Transfer learning techniques utilizing pre-trained models like VGG-16, ResNet-50, and MobileNet further enhance classification accuracy and reduce training overhead. The application, hosted on Krishi Megh, a cloud computing platform, offers real-time inference and expert-validated disease advisory through a user-friendly Android interface. Designed for use by farmers, veterinarians, and field extension workers, AI-DISA facilitates image capture or upload, automated disease identification, and delivery of species-specific management recommendations. It currently supports disease detection in bovines (e.g., mastitis, FMD, LSD) and canines (e.g., parvovirus, mange, tumors). This chapter details the end-to-end development pipeline of AI-DISA, including data collection, image annotation, model training, deployment architecture, and mobile integration. While highlighting its success, it also addresses current limitations such as limited species and disease coverage. Future directions include expanding the disease database, integrating with national surveillance systems, and improving model interpretability. AI-DISA exemplifies the transformative role of AI in veterinary diagnostics and its potential to enhance livestock health outcomes in resource-constrained settings.