Poubelles Intelligentes pour la Reconnaissance des Bouteilles et la Récompense des Utilisateurs

Abstract

Given the sharp increase in plastic waste and its negative impact on the environment, it has become essential to develop innovative and effective solutions to address this issue. In this context, this project leverages artificial intelligence technologies to design a smart and portable system capable of automatically detecting and sorting plastic bottles. The system is built around a Raspberry Pi unit integrated with a digital camera and various electronic components such as an electric motor, thermal printer, sensors, and indicator lights. For each bottle inserted, the system captures an image and analyzes it using an AI model trained on the MobileNet neural network — a lightweight and efficient model designed for low-resource embedded devices. A large and balanced dataset of over 70,000 images was collected and used to train and fine-tune the model. After integrating the model into the embedded system, experimental results demonstrated high accuracy and fast performance, making this solution a practical step toward enhancing automated sorting operations in various locations such as public areas or institutions, while also promoting environmental awareness and encouraging users to participate in recycling efforts.