ElectroSortNet: A Novel CNN Approach for E-Waste Classification and IoT-Driven Separation System

Recycling e-waste is paramount important because of its environmental impact and the valuable resources that can be recovered. According to the United Nations Environment Programme, there is 100 times more gold in a tonne of e-waste than in a tonne of gold ore. Therefore, this paper proposes a novel Convolutional Neural Network (CNN) architecture, named ElectroSortNet, for e-waste classification. The channel-based attention mechanism is provided by squeeze and excitation networks that provide improved channel dependency and access to global contextual information. Additionally, residual shortcut connections were incorporated to address the challenge of vanishing gradients in deeper networks. The proposed ElectroSortNet achieves a 97.44 % test accuracy over a novel dataset and outperforms several previous architectures. This paper also proposed an IoT-based end-to-end e-waste separation system using ElectroSortNet as the backbone, which incorporates a conveyor belt mechanism and robotic arms that separate the e-waste and non-e-waste based on the classification result of ElectroSortNet.