Skip to content

Production and performance

Automated counting

A camera over the belt counts every piece and records when product stops flowing.

How a camera countsEach piece is detected, tracked and counted when it crosses the lineillustrative example
Simulation: detected pieces with their IDs crossing the counting line
Illustrative simulation
  1. 01 Detects each piece
  2. 02 Tracks it with an ID
  3. 03 Counts on crossing

The problem

Counting by hand is slow, costly and rarely adds up

When a case holds too many or too few, the error is found late: at the scale, at the customer or in a complaint. Fixing it can mean reopening cases, recounting and repacking by hand.

In validation on the plant floor

Sound familiar?

  • Cases or bags with too many or too few units.
  • People assigned to counting instead of producing.
  • Differences between production counts and inventory counts.
  • Product given away because nobody knows how much actually went in.

What we ask first

Before proposing anything, we understand what happens on your line today.

  • What happens when a case holds too many or too few? Where is it found?
  • Does fixing the error mean reopening, recounting or repacking?
  • Who finds out about the error and who fixes it?
  • Where are damaged pieces detected and where do they go?

How we approach it

We install a camera where the count matters. A computer vision model detects and tracks each piece, counts it as it crosses a virtual line, and sends counts and events to the software, not the video.

  1. 1. We record and study your product

    We record real footage of the belt: speed, spacing between pieces, lighting and overlapping product.

  2. 2. We train and evaluate the model

    The model's count is compared against manual counts on clips the model has not seen.

  3. 3. We process on site

    An edge computer next to the line analyses the video; only counts and events go up.

  4. 4. We connect the count to the order

    Each count is linked to the line, product and order being run.

What it takes

Software

  • Detection and tracking model trained on your product
  • Count and event log by line
  • Count view in the production dashboard
The full platform →

Plant equipment

  • Industrial camera with a lens suited to the distance and belt width
  • Controlled lighting when plant light varies
  • Edge computer (Jetson-class) next to the line
  • Enclosure, mount, PoE switch and backup power (UPS)
How we select hardware →

Camera, lens and lighting are chosen for the product, the belt and the environment (cold, humidity, washdown); the final configuration is defined during technical design.

Who uses the information

Packaging operator
Sees the count for the current case or batch.
Line supervisor
Spots when pace drops or product stops flowing.
Inventory and shipping
Receives a produced quantity that matches what was packed.

Decisions it supports

  1. 01Stop or adjust filling before completing off-count cases.
  2. 02Free up people who count by hand today for other work.
  3. 03Investigate a shift or product where the count doesn't add up.

How value is measured

We measure before and after, on your line. Value is shown with your data, not promises.

Difference between manual and automatic count
Counted in parallel during validation, over the same periods.
Cases reopened or repacked due to count errors
Recorded before and after installation.
Hours spent counting by hand
Compared per shift before and after.
Estimate the value of this problem

Let's start with one problem.

Tell us what's costing you most today. We review it with you, show how we'd solve it and estimate its value together.