EPID: Description of the Enfield PCB Inspection Dataset for Visual Defect Detection

We present EPID, the Enfield PCB Inspection Dataset, a high-resolution image collection designed for benchmarking visual defect detection systems in printed circuit boards (PCBs). The dataset consists of 446 annotated images split into two subsets: a primary training set of 345 images and a separate validation set of 101 images. All images depict progressive physical damage to components such as integrated circuits (ICs) and capacitors, supporting temporal modeling and low-data learning scenarios. Each component is manually labeled as defective, non-defective, or ignored. EPID enables research in object detection, neural architecture search, and robust model generalization for industrial inspection tasks.