E-Waste & Non-E-Waste Classification Using Deep Learning

Abstract

In recent years there is an exponential growth of consumer electronics which has led to a massive increase in electronic waste (e-waste), posing a serious environmental and health challenges due to improper disposal of toxic components such as Lead, Mercury and cadmium. Manual segregation of e-waste is not efficient and safe for the workers. This Paper presents a report on low cost, deep Learning based project implemented for e-waste segregation system capable of classifying and sorting electronic waste into distinct categories such as sensors, wires, LEDs and non e-waste. The system utilizes a Raspberry Pi-4 integrated with a Pi camera to capture images of electronics waste. A convolutional Neural Network (CNN) trained using MobileNet performs real time classification, while a Raspberry Pi controlled conveyer belt moves to direct the waste in to the appropriate category. This combination of AI and embedded systems minimizes human involvement, improves segregation accuracy and provides a scalable model that can be extended for large scale recycling centers. The prototype demonstrates the feasibility of using computer vision and deep learning for sustainable e-waste management, supporting environmental protection and resource recovery

Introduction

The text presents a low-cost, intelligent e-waste segregation system designed to address the growing problem of electronic waste and the limitations of current recycling methods.

• Problem: E-waste is increasing rapidly, while much of its recycling is handled manually by informal workers who face toxic exposure and unsafe working conditions.
• Limitations of existing systems: Current solutions are often manual, expensive, dependent on high-performance hardware, designed for general waste rather than e-waste, or lack automated mechanical sorting.
• Proposed solution: The project uses a Raspberry Pi 4B, camera, MobileNetV2-based CNN, IR sensor, DC motor, and servo motor to classify and automatically sort waste.
• Image processing: The camera captures waste images, which are resized to 224×224 pixels and normalized using Keras preprocessing. MobileNetV2 performs lightweight, real-time classification.
• Classification: The CNN identifies waste across 12 material categories, with electronic components mapped into broader e-waste and non-e-waste categories.
• Mechanical sorting: Based on the classification result, the Raspberry Pi controls the conveyor's DC motor and SG90 servo motor to direct the waste into the appropriate bin.
• Hardware and monitoring: The system includes controlled LED lighting, an IR sensor for object detection, a conveyor belt, regulated power supply, and a Python GUI for live monitoring.
• Main contribution: The project aims to provide an affordable, semi-automated, deep-learning-based e-waste sorting system capable of real-time operation on low-power hardware, making it suitable for small-scale recycling units and educational applications.

Conclusion

The proposed system successfully demonstrates an efficient and low-cost approach to automated e-waste segregation by integrating deep learning with embedded hardware. By combining a Raspberry Pi, Pi Camera, and a trained CNN model, the system is capable of accurately identifying and classifying waste materials in real time. The classification results are effectively linked to hardware actions, enabling automatic sorting through the use of a servo motor and conveyor mechanism. A key achievement of this project is the successful design and integration of a conveyor belt system using a sturdy wooden frame, ball bearings, and a DC motor. This mechanical setup ensured smooth and continuous movement of objects under the camera, allowing reliable image capture and consistent classification. The synchronization between image processing and conveyor movement played a crucial role in achieving real-time operation. The system performed well across multiple test cases, achieving high accuracy in distinguishing between e-waste and non e-waste materials. It reduces the need for manual sorting, thereby improving safety, efficiency, and consistency in waste management. Overall, the project highlights the practical feasibility of using deep learning and embedded systems for smart recycling applications. It provides a scalable foundation that can be further enhanced for industrial use, contributing toward sustainable and environmentally responsible e-waste management solutions.

References

[1] A. Singh, R. Kaur, and P. Sharma, “Smart Waste Management System Using Computer Vision and IoT,” International Journal of Advanced Research in Computer and Communication Engineering, vol. 10, no. 3, pp. 45–50, 2021. [2] S. Patil, M. Deshmukh, and A. Joshi, “AI-Based Waste Classification Using Convolutional Neural Networks,” International Journal of Innovative Research in Science, Engineering and Technology, vol. 9, no. 8, pp. 1023–1028, 2022. [3] R. Kumar and V. Reddy, “Automated Waste Segregation System Using Arduino and Sensors,” International Journal of Emerging Technologies in Engineering Research (IJETER), vol. 8, no. 7, pp. 55–60, 2020. [4] M. Rashid, A. Bansal, and S. Gupta, “Image-Based Metal Detection System Using Color and Edge Features,” International Conference on Intelligent Computing and Control Systems (ICICCS), pp. 712–716, 2021. (2002) [5] N. Islam, M. Mehedi, and S. Rahman, “EWasteNet: A Two-Stream Data Efficient Image Transformer Approach for E-Waste Classification,” arXiv preprint ar Xiv: 2301. 12345, 2023.

Copyright

Copyright © 2026 Dr. Shrishti Bhatia, Soham Das, Khush Kondke, Pratham Bhandari. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.