Foreign object detection: Datasets & Models

Manual foreign object inspection is slow, inconsistent, and fatiguing, human inspectors can only sample a fraction of output and miss small or fast-moving contaminants under production speed. Automating this with a vision model means every unit passing under a camera is checked with the same rigor, around the clock, without inspector fatigue or shift-to-shift variance. The result is fewer contaminated products escaping to customers, less scrap from batch-level rejections, and a documented inspection trail for compliance. For QC teams under pressure to reduce defect escape rates, this converts an intermittent manual check into continuous automated coverage.

Teams get consistent, high-throughput contaminant screening that runs on existing line cameras with no bespoke sensor hardware. The RF-DETR model localizes foreign objects precisely so reject-and-divert systems can act on specific detections, and the approach scales from a single inspection point to every line in a facility. Because it integrates with standard camera feeds and downstream PLC or alerting systems, deployment fits into current QC infrastructure rather than replacing it.

1. Upload an image: Drop any photo into the widget above to try foreign object detection for free, right on this page.
2. Preview instant results: The model detects metal shaving, plastic fragment, screw, glass shard, wood splinter 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.