Reducing Appliance E-waste by Generating Repair Schematics Directly from Photos With CNN-GCN

The repair of electronic appliances is often hindered by the lack of available circuit schematics, leading to unnecessary waste and environmental harm.While right-to-repair legislation has improved access to repair services, circuitlevel diagnostics remain challenging, contributing significantly to electronic waste (e-waste).This paper hypothesizes that circuit schematics can be generated from PCB images to assist in repair and introduces Appliance X-ray, an AI-driven system designed for this purpose.Inspired by medical X-rays that help doctors diagnose patients, Appliance X-ray extracts and reconstructs schematics from PCB images, making circuit structures more interpretable.The system employs a YOLOv5 convolutional neural network (CNN) to detect circuit components, followed by k-nearest neighbors regression to predict missing elements based on inferred circuit functionality.A novel graph convolutional network (GCN) is then used to analyze component relationships to reconstruct the schematic.Additionally, human-in-the-loop feedback refines model predictions, enhancing future iterations.Experimental results demonstrate the effectiveness of this dual CNN-GCN model in identifying components and inferring connections, while also contributing a novel, scalable dataset of circuit schematic graphs derived from both real and synthetic data to support future research in circuit analysis and repair automation.