Paper Title
Transfer Learning-Based Weed Classification and Advisory System Using Deep Learning and RAG-Based Chatbot

Abstract
The uncontrolled proliferation of weeds poses a significant threat to agricultural sustainability, leading to reduced crop yields and increased farming costs. To address this challenge, we propose a sophisticated hybrid deep learning framework for weed classification, leveraging an ensemble of pre-trained Convolutional Neural Networks (CNNs) such as VGG16, VGG19, DenseNet201, and Xception. These architectures serve as base models, further refined with additional CNN layers, Long Short-Term Memory (LSTM)-based classifications, and Lightweight Recurrent Neural Networks (LRNN)-based classifications to enhance spatial and sequential feature extraction. The model is trained on a modified dataset consisting of directories, images, and corresponding labels to accurately identify weeds under various agricultural conditions. Additionally, we introduce a Retrieval-Augmented Generation (RAG) chatbot that provides real-time insights into weed impact on crops, preventive measures, and optimal herbicide applications. By integrating high-accuracy weed classification with knowledge-driven AI, our approach facilitates intelligent, automated decision-making, advancing precision agriculture and sustainable crop management. The proposed model achieves an impressive accuracy of 96.8%, demonstrating its effectiveness in real-world agricultural scenarios. Keywords - Weed classification, deep learning, CNN, VGG16, VGG19, DenseNet201, Xception, LSTM, LRNN, RAG chatbot, herbicide application, precision farming, sustainable agriculture