Computer vision
Web app
Manufacturing

Anonymous

A computer vision system for a frozen dumpling factory: a camera above each line counts every piece, rejects defects and checks there are exactly 16 in each bag. The shift supervisor sees all three lines in real time, and the shift report builds itself.

CLIENT: Frozen food factory, Vinnytsia region
INDUSTRY: Food manufacturing
YEAR: 2026
TIMELINE: 12 weeks

vision.local

“Shift now” screen: camera feed with detections and line counters

The task

A factory with three forming lines makes 6–9 tonnes of product a day in two shifts. Output was measured by weight and spot-checked by eye, so short-filled bags and defects — torn, stuck-together and misshapen dumplings — were spotted only at the warehouse or after complaints from retail chains. Supervisors filled in shift reports by hand, and the numbers didn’t match between shifts.

They needed a system that sees every piece, rejects defects on its own and gives supervisors and technologists exact numbers without manual counting.

What we did

01
Counting and classification
A camera after the forming machine counts every dumpling and classifies it: good, torn, stuck together or misshapen. Defects are rejected automatically, and low-confidence pieces go to a manual check.
- Computer vision
- Classification
- Rejection
02
Bag check
Before sealing, the system checks there are exactly 16 pieces in the bag. A short-filled bag is pulled and topped up instead of going to a retailer.
- Packing control
- Counter
- No short fills
03
Three lines in real time
The supervisor sees speed, plan vs actual and hourly defect rate for each line and gets a phone alert when defects go above the norm. The line operator has a tablet with a big counter and a button to call the supervisor.
- Dashboard
- Alerts
- Operator tablet
04
Reports and analytics
The shift report is generated automatically. The technologist sees defects by line and shift over the week and how they depend on dough temperature and line speed.
- Automatic reports
- Analytics
- PDF export

01 · FOR THE SUPERVISOR

Every dumpling counted

01 · FOR THE SUPERVISOR

vision.local/lines

Three lines: speed, plan vs actual and hourly defects

Piece count check before the bag is sealed

02 · ON THE PHONE

Supervisor and owner always in the loop

01Defect alert
02Shift summary

NEXT PROJECTKreminFundraising website and CRM for a charity with transparent reportingView case

Have a similar task? Let’s forge it.

Discuss a project

The task

They needed a system that sees every piece, rejects defects on its own and gives supervisors and technologists exact numbers without manual counting.

What we did

Counting and classification

A camera after the forming machine counts every dumpling and classifies it: good, torn, stuck together or misshapen. Defects are rejected automatically, and low-confidence pieces go to a manual check.

Computer vision
Classification
Rejection

Bag check

Before sealing, the system checks there are exactly 16 pieces in the bag. A short-filled bag is pulled and topped up instead of going to a retailer.

Packing control
Counter
No short fills

Three lines in real time

The supervisor sees speed, plan vs actual and hourly defect rate for each line and gets a phone alert when defects go above the norm. The line operator has a tablet with a big counter and a button to call the supervisor.

Dashboard
Alerts
Operator tablet

Reports and analytics

The shift report is generated automatically. The technologist sees defects by line and shift over the week and how they depend on dough temperature and line speed.

Automatic reports
Analytics
PDF export