Design and Implementation of an AI-Powered Maize and Tomato Crop Disease Detector for Local Farmers

This project presents CropDiseaseDetector, an AI-powered offline diagnostic tool for maize and tomato crop diseases, developed for smallholder farmers in Northern Nigeria. The system uses a convolutional neural network trained on the Kaggle New Plant Disease dataset (14 disease classes) and deployed as a TensorFlow Lite model for on-device inference on an Orange Pi AIpro (Huawei Ascend NPU), enabling full offline functionality with no internet dependency. The application combines a Flask backend with a React frontend packaged in Electron (TypeScript), an SQLite database for local scan history, and a multilingual interface supporting English, Hausa, Yoruba, and Igbo. The system targets diagnostic accuracy exceeding 90% for major maize and tomato diseases, aiming to reduce diagnosis time from days to seconds compared to traditional laboratory and extension-service methods, and to make early, actionable crop disease intervention accessible to smallholder farmers regardless of connectivity or cost barriers.