AI-powered waste management system in Egypt for detecting, classifying, and rewarding garbage collection.
Naddafly is an AI-powered waste management mobile application built to tackle garbage problems in Egypt. It enables users to detect, classify, and report street garbage through an AI model trained on a custom dataset of 30,000+ labeled images. The app includes a reward-based system that motivates users to participate in cleaning efforts while helping collectors locate and manage garbage pickups efficiently.
Collected and annotated over 30,000 street garbage images across Egypt, then trained a YOLOv8 object detection model that achieved 90% accuracy in identifying garbage in real-world environments.
Developed a convolutional neural network (CNN) model to estimate garbage quantity (small, medium, or large) based on the detected image, providing accurate volume classification for better logistics.
Implemented the backend using Flask to manage API requests and AI inference. The mobile app, built in Kotlin with Android Studio, allows users to capture garbage images directly from their phone camera.
Introduced a gamified incentive system where users earn points or vouchers for successfully detecting and reporting garbage, redeemable at partner stores and restaurants.
Designed a dedicated interface for garbage collectors to view real-time detection points on Google Maps, plan efficient collection routes, and confirm successful pickups.
Used Label Studio for precise data annotation to build a high-quality dataset for model training, ensuring reliable performance across diverse urban scenarios.