TinyML driven real-time sustainable waste sorting using IoT
Sivakumar Premkumar, T.S. Arthi, Ritesh Kumar, Ronak Singh · 2025
Effective waste management is crucial for addressing environmental challenges, yet current waste sorting systems are often limited by high costs, energy inefficiency, and reliance on cloud-based processing. These limitations hinder the scalability and real-time application of automated sorting technologies. This paper explores the application of TinyML, a lightweight form of machine learning optimized for edge devices, in automating waste classification into categories such as plastic, glass, metal, organic, and paper. The primary motivation aims to reduce complexity and high energy requirements of conventional methods and design a scalable and energy efficient system that can be implemented using low power microcontrollers. Current frameworks are computational intense and rely on data transfer to the cloud, which results in latency and privacy issues. This work fills these gaps through on-site data processing with TinyML, which greatly minimises energy consumption and increases interactivity. The outcomes of this research show that the proposed system has the capabilities to classify the wastes with high accuracy, and at the same time may not involve many operational expenses. The research is unique and important because it offers an efficient, reliable and eco-friendly way of sorting wastes fully automated to aid in smart waste disposal.