A low-cost cross platform mobile application with machine learning features has been implemented using Flutter for users to capture diseased images of leaves and get the plant, disease name and cure for fungal diseases in Mauritius.
Agriculture plays a crucial role in the Mauritian economy; however, plant growth and crop yield can be affected by the infection of fungal diseases. Mauritius does not have an on-the-go solution for identifying and providing cures for such diseases, farmers mostly rely on their knowledge and experience. A low-cost cross platform mobile application with machine learning features has been implemented using Flutter for users to capture diseased images of leaves and get the plant, disease name and cure. In addition, the system uses the location of the captured images to provide historical weather data and allows users to contact FAREI (Food and Agriculture Research Extension Institute) for further details. Furthermore, image scans are stored in Firestore Database and a summary is provided. A chat assistance is also provided to communicate with an expert in the field with the ability to send/receive images. A CNN (Convolutional Neural Network) model has been trained with 3000 images in 25 epochs using Python on Kaggle, yielding an accuracy of 75.5%. An accuracy of 30.0% for Black Rot and 60.0% for Rust was achieved using a training set of 10 images per disease and an overall accuracy of 68.3% was recorded from a test audience of 60 people.