Ai Agri-Gaurd : Crop Disease Detection

Authors: Aashi Vidyarthi, Aadesh Rathore, Devika girase, prarthi patel

Abstract: Agriguard is an intelligent crop-disease detection and advisory platform that leverages advanced technologies from the fields of Artificial Intelligence (AI), Computer Vision, and Deep Learning to provide real-time plant health monitoring. It automates the identification of crop diseases by analyzing digital images of plant leaves and generates immediate treatment recommendations. At its core, the system employs Deep Learning, a subfield of AI that enables computers to automatically learn features and patterns from data without explicit programming. Deep learning uses artificial neural networks inspired by the human brain, where interconnected layers of neurons process input data and progressively learn complex representations. In Agriguard, the primary model used is ResNet-50, a 50-layer Convolutional Neural Network (CNN) architecture. CNNs are a type of deep neural network particularly effective for image classification tasks because they automatically detect spatial features such as edges, textures, and shapes. Unlike traditional machine learning methods that depend on handcrafted features, CNNs perform hierarchical feature extraction — learning low-level details (like leaf color or vein texture) in early layers and more abstract disease-related patterns in deeper layers. The ResNet (Residual Network) architecture introduces the concept of skip connections or residual links, which allow gradients to flow directly through layers during backpropagation. This innovation prevents the “vanishing gradient” problem and makes it possible to train very deep networks effectively. ResNet-50, pre-trained on the large ImageNet dataset, has been fine-tuned with agricultural datasets, including PlantVillage and region-specific field images. This transfer learning approach significantly improves model accuracy and reduces training time. The trained model achieves over 94% classification accuracy across multiple crop types, successfully differentiating between diseases such as early blight, late blight, leaf mold, and healthy leaf conditions. The trained model is deployed through a Flask-based REST API, which acts as a communication bridge between the AI model and the user interfaces. Flask, a Python micro web framework, allows efficient server-side handling of requests where an image uploaded by the user is processed by the backend model to generate a prediction. The REST (Representational State Transfer) API ensures seamless communication using standard HTTP methods, making the system modular and scalable. The front-end of Agriguard is built using React, a modern JavaScript framework known for its component-based architecture and high responsiveness. The React-based web dashboard and mobile application allow users— especially farmers—to capture or upload images of affected leaves through their devices. Once processed, the system displays the disease name, probability score (confidence level), and recommended treatment measures, including pesticide usage, organic remedies, and preventive steps to avoid recurrence. Agriguard’s goal is to make precision agriculture—a data-driven approach that optimizes inputs like water, fertilizer, and pesticides—accessible to every farmer, regardless of technical expertise.