An AI-Enabled Multimodal System Combining Visual Leaf Data and Multilingual Symptom Inputs for Disease Prediction and Treatment Recommendations in Rice and Wheat

Artificial intelligence techniques, you know, especially those deep learning algorithms, have really picked up in the farming sector lately. That sector is just one of many getting hit by this stuff. With deep convolutional neural networks, or CNNs like they call them, things are changing fast. supported by multimodal input handling, we propose a new approach to the study of crop diseases in this research. The proposed system here will provide an effective, cost-saving and affordable way of remote detection of disease in rice and wheat crops. We employ an optimized convolutional neural network (CNN) design for leaf image processing combined with analysis of multilingual symptom inputs. Furthermore, we also use data augmentation techniques strengthen the model in terms of its generalization ability and efficiency in handling various crop conditions and disease patterns. Our approach consists of advanced image preprocessing, segmentation by enhanced feature extraction methods, classification by Deep Convolutional Neural Networks (DCNN) with multilingual natural language processing capabilities, and real-time treatment recommendation processes which improves the accuracy and reliability of the system to identify crop diseases based on visual and textual symptoms. The agri-advisory platform achieves strong performance with ease of accessibility through its voice/text interface multilinguality and location-based service integration utilizing React.js and Node.js for frontend, backend with Express.js and MongoDB for database system. We have deployed the backend in Node.js using Express, and TensorFlow.js has been utilized in powering the inference engine. All model checkpoints and media assets are stored in Google Cloud Storage buckets, and MongoDB Atlas handles user profiles, disease logs, and treatment records. The system incorporates real-time chatbot features for interactive QandA assistance and offers location-based suggestions for local pesticide and fertilizer shops facilitating farmers to act at the right time.