REAL-TIME COTTON LEAF DISEASE DETECTION AND PESTICIDE RECOMMENDATION USING DEEP LEARNING AND CONVERSATIONAL AI

Abstract

The increasing susceptibility of cotton crops to various diseases significantly affects agricultural productivity, particularly in regions where cotton is a major commercial crop. This project presents a real-time cotton leaf disease detection and pesticide recommendation system that integrates deep learning and conversational AI to support farmers in early diagnosis and effective decision-making. The proposed system utilizes a two-stage approach, where a YOLO-based
model detects infected regions in cotton leaf images, followed by an EfficientNet-
based classifier that identifies specific disease categories with high accuracy.
In addition to disease classification, the system incorporates confidence-based decision logic to handle uncertain predictions and enhance reliability in real-world conditions. A curated knowledge base provides appropriate pesticide recommendations, including dosage, application guidelines, and safety measures. Furthermore, a multilingual chatbot interface enables farmers to interact with the system and receive context-aware advisory support in an accessible manner.