Generalization of Missing Component Detection Across Diverse PCB Layouts

Establish reliable missing component detection on diverse printed circuit board layouts, overcoming the poor generalization of current supervised object detectors and unsupervised anomaly detection methods to unseen board designs.

Background

PCB-MC introduces footprint-level annotations for components that are absent from printed circuit boards and evaluates both supervised object detectors and unsupervised anomaly detection methods. The benchmark covers 197 distinct board types and uses board-type-aware cross-validation to prevent models from exploiting repeated layouts or augmented samples.

The reported results show high false-negative rates for supervised detectors on unseen board designs, while unsupervised anomaly detection methods fail because they lack spatial alignment with a board-specific reference. The authors therefore identify reliable missing component detection across diverse PCB layouts as an unresolved challenge requiring further methodological development.