Agro-Doc: Design and Implementation of an AI-Powered Web Platform for Plant Disease Detection

Plant diseases significantly impede agricultural pro- ductivity and global food security. Conventional diagnostic meth- ods are often slow, costly, and inaccessible, particularly for farmers in remote areas. This paper presents Agro-Doc, a web- based platform employing artificial intelligence (AI) for rapid and accessible plant disease identification. Agro-Doc utilizes a Convolutional Neural Network (CNN) based on a modified ResNet-50 architecture, enhanced with transfer learning, to analyze plant images submitted by users via a web interface. Key features include real-time disease classification, detailed disease information, treatment suggestions, and a community forum for knowledge exchange. A notable aspect is the deployment of the core AI model using TensorFlow.js, enabling efficient inference directly on the client-side (user’s browser), thereby enhancing speed, privacy, and offline potential. The system, built with Re- act.js, Node.js, and MongoDB, demonstrated an overall accuracy of 94.2% across 42 diseases and 15 crops in validation tests. Field condition accuracy was 88.5%. User evaluations confirmed high usability (average SUS score: 82.1). Agro-Doc showcases a practical approach to democratizing AI-driven agricultural diagnostics through accessible web technologies.