Updated July 2026 · 6 min read · Adente Vision Engineering Team
What is the difference between machine vision, computer vision and AI vision?
Buyers use the three terms as if they were one, and a vendor call gets confused within a minute. They are not synonyms, they are nested. Computer vision is the parent field, machine vision is its industrial application, and AI vision is the newer layer where the decision comes from a trained model instead of a fixed rule. Getting the words straight is the fastest way to compare two datasheets that describe the same task in different language.
| Term | What it means | On the factory floor |
|---|---|---|
| Computer vision | The broad field: any software that extracts meaning from images | Research and algorithms, the parent discipline behind the rest |
| Machine vision | Computer vision applied to an industrial task, historically rule-based | A camera plus fixed rules reading a barcode or gauging a dimension |
| AI vision | Machine vision where a trained model, not a rule, makes the judgement | Learns a good part from examples, flags anomalies, classifies variants |
| Hybrid vision | Classical vision for measurement combined with AI for judgement | Measures in millimetres with classical vision, judges defects with AI |
The rows nest downward. Everything in the machine-vision row is also computer vision; everything in the AI-vision row is also machine vision. What changes as you move down is where the decision comes from: a human-written rule at the top, a learned model at the bottom, and a deliberate combination of both in the hybrid row.
Why did "machine vision" come to mean rule-based?
Machine vision earned its rule-based reputation because that is what the technology was for thirty years. Before deep learning was practical on a factory budget, an industrial camera fed fixed algorithms: find this edge, measure this distance, read this code, compare this pixel value to a threshold. The logic was explicit, so it was fast, deterministic and easy to validate, and it still is the right tool for a barcode or a fixed gauge.
That history is why "machine vision" and "rule-based" are still treated as the same thing on many sites, and why "AI vision" is sold as a separate category. On the floor the distinction is simpler than the marketing: a rule-based check can only find a defect someone described in advance, while an AI check learns the appearance of good or bad from examples and generalises to cases no one wrote down. The camera and lighting can be identical; what differs is who makes the call, a rule or a model. For the practical decision of when to reach for one over the other, see rule-based vs AI machine vision.
Where does hybrid vision fit between the two?
Hybrid vision is the deliberate choice not to pick a side. Classical computer vision is unbeaten at measurable geometry: it locates a feature and measures a distance in millimetres with repeatable precision. AI inference is unbeaten at judgement: it recognises a variant or flags a surface flaw that no threshold could describe. A hybrid unit runs both, using each where it is strongest, so the result is auditable rather than a single black-box verdict.
Adente Vision is an edge-AI visual inspection unit built by ADENTE Advanced Engineering Technologies, part of the Aden Group, sold through automation system integrators. It combines classical computer vision with AI inference and exposes four inspection modes, Anomaly, Defect, Counting and Quality, so one unit measures a dimension with classical vision and judges a rare defect with AI on the same part. That pairing is why "AI vision" on its own is an incomplete label for what an industrial unit actually does.
Which term maps to which datasheet capability?
Once the terms are clear, a datasheet is easier to read, because each capability belongs to one layer. The imaging front end, the camera and lighting, is classic machine-vision hardware, and it is where numbers like resolution and shutter type live: an up-to 12 MP global-shutter camera with configurable lighting captures the image regardless of which layer decides on it. Global shutter matters because it freezes a moving part without smear, so the picture the model sees is a true one.
The decision layer is where "AI vision" is earned. On-device inference on a Jetson-class edge board, running the four inspection modes, is the part that learns a good part from about 20 images and scores a new one against it. Integration is a third, separate concern: pass/fail and program-select travel over PROFINET, EtherNet/IP, Modbus TCP, EtherCAT or OPC UA and 4 discrete inputs and 4 outputs at 24V, no matter whether a rule or a model made the decision. Reading a spec by these three layers, imaging, decision, integration, tells you which term a vendor is really selling.
This post is a spoke of the pillar guide on AI visual inspection; to see the imaging, decision and integration layers in one unit, take the system overview.