What is computer vision?

Computer vision is AI that interprets images and video. IBM describes it as a field that "equips machines with the ability to process, analyze and interpret visual inputs such as images and videos". The common tasks are classifying an image, detecting and locating objects within it, and measuring what it finds.

In a timber business

Vision systems count and measure logs on trucks and in yards, grade logs and boards, find knots, cracks, wane and rot, and check finished products. The shift the research describes is to deep learning: models that, in the words of the Sensors study below, "automatically detect and extract high-level image features from labelled image data" rather than following hand-written rules.

Detection is not value

Many papers report how well a model detects things, as a precision or accuracy score on a test set. The study below is typical: its best model scored a mean average precision of 0.94. A plant cares about something else: what the cuts or grades that the model drives are worth.

A 2024 study in Sensors did test the second. Its model located knots, pith and cracks on sawn timber and guided where to cut. On 100 pieces, yield rose from 69.2% with fixed-length cutting to 81.5%, a gain of 12.3 points. That is a real gain, but against the simplest baseline. A plant that already cuts with an optimiser starts higher, so we would expect the gain against its current method to be smaller.

The same study listed the kinds of errors that matter: one live knot "missed due to its excessively light color", and "a shadow along the edge of the lumber" labelled as a dead knot. Each moves a saw cut.

What it isn't

Computer vision isn't the camera, which only captures the image; the model does the interpreting. Nor is it infallible: in the study above, pale colour, shadows and dense grain were all misread.

What to check

Test on your own material, under your own lighting. Detection scores help compare models, but judge the whole system by its effect on yield or value against your current method. Count errors by what they cost.

Quarri for sawmills is built around how a sawmill runs, from log intake to shipped order.

Quarri is an AI-native data platform for the timber supply chain. It connects buying, production, sales and inventory for forest management, sawmill, wood products and pulp, paper and packaging operators.