Trinocular AI · Production Intelligence

Production Monitoring

Visual inspection on the line camera you already have, catching a defect at the station that produced it rather than at dispatch.

See the Demo
A factory conveyor camera view. Trinocular AI has drawn a detection box around each panel passing the inspection station, labelled Tile OK at 98 per cent, while a worker in a hi-vis vest and helmet looks on.
CAM 12 · Line 2 · 09:06:41 Tile OK 0.98

What it does

Catching it at the station, not the gate

A defect found at dispatch has already consumed every downstream operation it passed through. Found at the station that produced it, the same defect costs one item and a correction. Continuous inspection on the line is what moves the detection point upstream.

What gets checked

Each item passing the inspection point is checked for surface defects, edge and finish quality against the criteria you define as acceptable. Throughput and reject rate are counted by line and by shift, so the numbers a supervisor reviews are measured rather than sampled.

Drift before it becomes a batch

The value is often less in any single reject and more in the trend. A slow rise in a particular defect type on a particular station is the signal that something has moved out of tolerance, and seeing it in the first dozen items rather than the first thousand is what prevents a batch write off.

Honest limits

What can be judged depends on resolution, lighting and the angle the camera is mounted at. A defect on a face the camera cannot see will not be caught, and some surface finishes need controlled lighting to assess reliably. We assess the specific line position before committing to what the model can judge.

At a glance

Inspects
Surface defects, edges, finish
Counts
Throughput and reject rate per shift
Flags
Station and timestamp per reject
Deployment
Edge inference on the line
  • Inspect Surface defects, edges and finish per item
  • Count Throughput and reject rate by line and shift
  • Flag Station and timestamp for every reject
  • Trend Drift caught before it becomes a batch

Runs on the line camera you already have. What can be judged depends on resolution, lighting and how visible the defect is from the mounted angle.

Questions

Do we need a special inspection camera?

Not always. Many lines already have a camera positioned usefully. Where the existing position cannot see the surface that matters, or the lighting is not controlled enough to judge finish, we will say so rather than deploy something that reports confidently and wrongly.

Can it integrate with our MES or quality system?

Detections and counts are available as events, so reject records and throughput figures can be pushed into the system your quality team already works in rather than living in a separate dashboard.

Get Started

See it on your own cameras

Bring your existing CCTV. We will assess what your angles can actually support and show you the detections running on your own footage.