A Low-Resolution Used Electronic Parts Image Dataset for Sorting Application

Abstract

The accumulation of electronic waste (e-waste) is becoming a problem in society. Old parts and components are conveniently discarded instead of being recycled. Economic and environmental measures should be taken by individuals and organizations to enhance sustainability. This could include desoldering and reusing parts from electronic circuit boards. Hence, the purpose of the dataset presented in this paper is for the classification of used electronic parts in linear voltage regulator power supply circuits. The dataset presented in this paper comprises low-resolution (30 × 30 pixels) grayscale images of major reusable electronic parts from a typical adjustable regulated linear voltage power supply kitset. The three major reusable parts are capacitors, potentiometers, and voltage regulator ICs. These are typically the most relatively expensive components. Data representing the parts are extracted from 960 × 720 pixel workspace images containing multiple parts. This permits the dataset to be used with multiple types of classifiers, such as lightweight shallow neural networks (SNNs), support vector machines (SVMs), or convolutional neural networks (CNNs). Classification accuracies of 93.5%, 94.9%, and 98.4% were achieved with SNNs, SVMs, and CNNs, respectively. Successful detection and classification of parts will permit a Niryo Ned robotic arm to pick and place parts in the desired locations. The dataset can be used by other academics and researchers working with the Niryo Ned robot and Matlab to handle electronic parts. It can be expanded to include relatively expensive components from other types of electronic circuit boards.