Demeter - The Farming Assistant

Abstract: India's agri-sector accounts for 18% of GDP but suffers from systemic inefficiencies such as delayed disease detection (accounting for 30% annual losses), fertilizer overuse, and inaccessibility of experts. Demeter is a system combining computer vision (CNN), ensemble learning (Random Forest), and NLP to offer: (1) Real-time disease identification through leaf image analysis (98.7% accuracy), (2) Data-based crop/fertilizer suggestions (97.3% accuracy), and (3) Multilingual chatbot support. The following paper elaborates on system architecture, ML model training, and field trial outcomes from over 250 farmers in Maharashtra.