PCB defect detection: Datasets & Models

Manual PCB inspection is slow, inconsistent, and hard to scale as board density and production volume grow, and human inspectors miss subtle defects like hairline solder bridges or partial opens after hours on the line. Automating pcb defect detection with computer vision gives you repeatable, high-speed inspection that catches these faults consistently across every shift. In a manufacturing and QC setting, catching defects at the board stage prevents costly downstream rework and field failures. The result is faster throughput and a documented, auditable record of every inspected board.

Teams get consistent defect classification and localization that runs at line speed and scales across multiple inspection stations without adding headcount. The workflow integrates with existing inline cameras and inspection fixtures, so you can deploy on current infrastructure rather than re-tooling. RF-DETR provides strong accuracy on small, densely packed defects, making it a solid recommended starting point for building your own PCB inspection pipeline.

1. Upload an image: Drop any photo into the widget above to try pcb defect detection for free, right on this page.
2. Preview instant results: The model detects missing hole, mouse bite, open circuit, short circuit, spurious copper, spur within seconds and returns an annotated image with structured predictions, so you can immediately see how well it works. Adjust the detection classes and re-run until the results fit your scenario.
3. Like what you see?: With a free Roboflow account, you can deploy this workflow with the Serverless Cloud API, customize it in your workspace, or train RF-DETR on your own images for production accuracy.
4. Get started for free: No code or training data required to begin: upload, review the results, and make it part of your stack in a single click.