Cropsense-Crop Disease Detection and Alert Generation using SAM2

Abstract

M. Tejaswi, K. Srija, K. Teja, L. Harika and M. Yuvaraj

Agriculture continues to suffer major losses due to late and inaccurate detection of crop diseases. Manual monitoring is slow, error-prone, and impractical for large-scale farms. To address these challenges, Crop Sense introduces an intelligent platform for crop disease detection and management. Using Segment Anything Model 2 (SAM2) for precise visual segmentation and a ResNet50 CNN classifier trained on 65 disease classes across 13 crop types — including Rice, Cotton, Sugarcane, and Wheat — the system identifies disease type and assesses severity with 98.69% validation accuracy. Beyond detection, Crop Sense introduces three novel capabilities: (1) a weather-integrated disease spread prediction engine that forecasts disease progression risk over 7 days using real-time Open Weather Map data; (2) a drone video analysis module that processes aerial footage frame-by-frame using SAM2 segmentation to generate field-level disease maps; and (3) an AI-powered multilingual farmer chatbot powered by Llama 3.2 that provides crop disease diagnosis in Telugu, Hindi, and English. The platform provides comprehensive digital prescriptions covering biological remedies, chemical treatments, and preventive measures tailored to each specific disease and its severity intensity.

PDF