Visible defects, caught on the line.
Detect visible product, surface and packaging defects with models trained on your own good and bad samples — with an image saved for every reject.
Inspection that does not get tired.
Manual inspection misses defects on long shifts. MachVizyon trains a detection model on your own product images and runs it on a camera at the line, flagging defects as parts pass.
Every flagged part is saved with its image, defect class and time, so quality teams can review, retrain and trace issues back to their source.
Custom-trained models
Built on your own good and defective samples, not generic datasets.
Surface defects
Cracks, scratches, dents and marks visible to the camera.
Assembly checks
Missing, misplaced or wrong components.
Packaging & seals
Damaged packaging, open seals and print issues.
Image evidence
An image saved for every reject, ready for review.
Improves over time
New samples from the line feed back into retraining.
From frame to decision.
Capture
A line camera captures each part or product.
Detect
The model finds and classifies visible defects.
Decide
Rules decide pass or reject per defect class.
Signal
Rejects raise an alert or a signal to your line.
Log
Image, class and time are stored for quality review.
Where it fits
- Metal and machined parts
- Packaging and labelling lines
- Assembly verification
- Any defect a person can see on camera
What you need
- A camera with a stable view of the product
- Consistent lighting at the inspection point
- Sample images of good and defective parts
- An NVIDIA Jetson edge device or GPU server
Show us the defect. We’ll show you the model.
Book a site survey and a live demo on your own CCTV.