Code for Agriculture 1st Place

GreenScreen | Ethiopia

Plant diseases, pests, and unwanted weeds can damage crops, compete for nutrients, reduce agricultural productivity, and cause significant losses, yet identifying these problems is not always easy for farmers, gardeners, and plant owners who may not have immediate access to agricultural experts. GreenScan is an AI-powered plant diagnosis and identification system designed to make identifying plant diseases, pests, and weeds faster, simpler, and more accessible. The system allows users to provide a photograph of an affected plant, describe its characteristics or symptoms using text, or use both forms of input to receive a possible identification. At the center of GreenScan is an AI recognition system trained using approximately 1,700 plant images to recognize more than 51 plant diseases. The system uses three complementary recognition approaches: image-based recognition, which analyzes visual characteristics from an uploaded photograph; text-based recognition, which evaluates descriptions of observable symptoms or plant characteristics; and combined recognition, which uses both image and text information when available. These approaches allow GreenScan to accommodate users who have a photograph, users who can only describe what they see, and users who can provide both. The platform is organized into three identification modes: Disease, Pest, and Weed. In Disease mode, a user can upload a photograph or enter a symptom description, and the system processes the input to identify the most likely disease. In Pest mode, users can investigate harmful organisms affecting their plants. In Weed mode, GreenScan helps users identify unwanted plants growing among or around their intended crops or garden plants, allowing them to distinguish potentially invasive or competing vegetation from the plants they want to grow. After processing the input, GreenScan presents a possible identification together with relevant information and recommendations intended to help users understand the problem and consider appropriate next steps. The system is designed to make the initial identification process more accessible without requiring users to possess specialized plant pathology or botanical knowledge. Its software-based structure also creates opportunities for future expansion, including additional crops, diseases, pests, and weeds, more regional training data, local-language support, offline or low-connectivity functionality, and more localized agricultural guidance. GreenScan is not intended to replace agricultural experts, but to provide an accessible first layer of plant-health assistance that can help users recognize potential problems and make more informed decisions. By combining AI-based image and text recognition with a practical, user-friendly interface, GreenScan aims to bring useful plant-identification information closer to the people who need it.