PranaDhanya: An Intelligent Portal for Crop Disease Diagnosis, Livestock Health Screening and Farm Decision Support

Small and marginal farmers usually have to decide about crop disease, livestock illness, fertilizer and selling time without quick access to an expert. This paper presents PranaD-hanya, a web portal that places six machine-learning models behind one role-based interface for farmers, veterinary offi-cers and administrators. Two ResNet9 convolutional networks screen leaf images (38 plant classes) and animal images (11 cat-tle, dog and goat classes); an XGBoost classifier recommends crops from soil and weather values; Random Forest models es-timate crop yield, wholesale price and a suitable fertilizer. A FastAPI microservice serves the models to a Spring Boot back-end secured with JSON Web Tokens and backed by a Post-greSQL database. On held-out data the plant and animal im-age models reach 96.00% test accuracy, crop recommendation reaches 98.64–99.86% accuracy, fertilizer prediction reaches 98.99–100%, and the yield and price regressors reach R2 of 0.74–0.85 and 0.77–0.85. We also discuss where the system still falls short, especially for field photographs taken against cluttered backgrounds