Pothole detection: Datasets & Models

Manual road condition surveys are slow, subjective, and expensive: crews drive routes and log defects by eye, or citizens report potholes long after they form. Automating pothole detection with computer vision lets a municipality or contractor scan miles of roadway from ordinary vehicle cameras and generate consistent, geotagged damage inventories without adding inspection staff. For construction and infrastructure teams, this means faster prioritization of repairs, better budgeting from objective severity data, and the ability to re-survey the same roads over time to track deterioration.

Detection runs on footage from cameras you already deploy on fleet vehicles or inspection drones, so there is no specialized hardware requirement to start. The workflow scales from a single maintenance truck to a citywide network, producing repeatable, quantifiable defect counts instead of hand-written notes. Training on RF-DETR gives strong accuracy on small and partially occluded road defects while remaining practical to deploy at the edge or in the cloud.

1. Upload an image: Drop any photo into the widget above to try pothole detection for free, right on this page.
2. Preview instant results: The model detects pothole, crack, manhole cover, road patch, speed bump 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.