Solutions

Automated defect detection with vision AI

Train a model on images of your own parts to catch cracks, dents, missing components, and contamination on every unit. Roboflow runs on the cameras you already have and sends pass or fail to your line in real time.

A machined aluminum part on a conveyor with a crack detected at 97% confidence, shown magnified in an inset

What is automated defect detection?

Automated defect detection uses cameras and vision AI to find flaws in products as they move through production. A model trained on your own images inspects every unit, instead of a sample from each lot, and triggers an action the moment it finds a defect.

  • A laptop lid on a roller conveyor with a scratch detected at 96% confidence and magnified four times

    Surface defect detection

    Catch scratches, cracks, and abrasions at line speed, including the kind that only catch the light at a specific angle, before units ship.

  • A white sheet-metal panel under inspection lights with a 0.4 mm dent detected at 98% confidence

    Dents and deformation

    Flag dents, warping, and dimensional deviation on formed and machined parts as they occur, so damaged goods never reach your customers.

  • A bolted flange seen from above: seven fasteners pass and one missing fastener is flagged at 98% confidence

    Missing components and assembly errors

    Verify every build against the expected configuration: missing fasteners, wrong parts, and unseated connectors caught before the next station.

  • Trays of rice on a food production line with a foreign object detected at 97% confidence in one tray

    Contamination and foreign objects

    Detect contamination and foreign objects before sealing and shipping, and stop a single missed inclusion from becoming a recall.

How it works

Train an accurate defect detection model with less data

You don't need millions of defect images to start. Here's how a model goes from your first labeled images to inspecting every unit.

  • Capture

    Use the cameras on your line

    Roboflow works with the GigE and USB industrial cameras, smart cameras, and edge devices you already have, with no line redesign.

  • Label

    Label defects in minutes

    Describe a defect in plain language and Auto Label marks it across your images. Your quality team reviews every label before training.

    Explore Auto Label
  • Train

    Train on your own parts

    Train a model on images of your own parts and defects, then check its accuracy class by class before it goes near the line.

    Explore Train
  • Deploy

    Inspect every unit at the line

    Run the model on an edge device next to the camera, whether your site is connected, air-gapped, or fully on-prem.

    Explore Edge
  • Improve

    Get better after deployment

    Active learning collects the images your model finds hardest. Your team labels them, and each retrain adds what the line has taught it.

Act on every defect in real time

A detection is only useful if the line reacts. Roboflow Workflows connects each result to the equipment and people on your line.

  • Send pass or fail to your PLC to drive reject gates, marking systems, and line stops.
  • Alert operators the moment a defect appears.
  • Log every detection with its image, timestamp, and result, supporting the records your ISO 9001, IATF 16949, and FDA 21 CFR Part 11 audits expect.
A defect detected on the line and the part sent to a reject gate

Inspection runs on site in all three network modes.

  • Connected

    Default

    Devices reach Roboflow Cloud over one outbound connection.

    Training
    Roboflow Cloud
    How models arrive
    Over the air, and Secure Gateway can cache them for the whole site.
    Internet
    Outbound on port 443 only, with no inbound ports.
    Best for
    Most plants.
  • Air-gapped

    Devices run on a closed network once the model is on them.

    Training
    Roboflow Cloud
    How models arrive
    Downloaded once, then cached on the device.
    Internet
    Only to download or refresh a model.
    Best for
    Sites whose devices can’t reach the internet, or can’t count on it.
  • Fully on-prem

    Regulated sites only

    Roboflow runs on your own servers, with no cloud at all.

    Training
    Your own servers
    How models arrive
    From your Roboflow install, over your own network.
    Internet
    None.
    Best for
    Regulated sites that can’t train models in the cloud.
Start where you are

Speak with an AI expert

Our team will help you start solving business problems on the first call.

  • Solution architecting
  • Live demonstration
  • Pricing and specifications
  • Feasibility assessment

Over 80,000 organizations build with Roboflow.

Rivian
Pella
Chobani
USG Corporation
BNSF Railway
American Woodmark
Outcomes
Patrick Industries
Peer Robotics