Check visible quality features
We test whether images can reveal features such as completeness, position or a defined defect. Lighting and product variations are included in the assessment.
AI video analytics on existing cameras The camera spots the agreed event and sends a snapshot for review.
A part goes missing from the line, someone blocks a walkway, items need counting: we check whether your cameras can handle it.
One concrete process is enough to get started.
FROM NEED TO IMPLEMENTATION
- 01 · Check the view and the taskWe review approved footage or the site and establish the camera angle, image quality, event definition and handling of recordings.
- 02 · Test in working conditionsThe test covers typical operating conditions and difficult situations. Detected and missed events are compared with a manual review.
- 03 · Plan the response and operationYou receive the findings and a proposal for alerts, responsibilities, required equipment and further testing.
Real camera footage from our projects in a plant and a kitchen.
VIDEO ANALYTICS IN UNDER TWO MINUTESCameras that count.
Cameras that count.
And alert.
Car washes, plant, kitchen and entrance: what video analytics detects in our projects. Plus retail: how a shoplifting alert is raised.
PROJECT SCOPE
What we build together.
Monitor events and areas
Count items, identify occupancy or flag entry into a defined area. Together, we specify an event that can be clearly observed.
Evaluate the alerts
The pilot report documents detected events, false alerts and missed cases using reviewed footage. This helps you assess practical value.
FROM PRACTICE
Count vehicles. See the load.
A system in daily operation at a chain of car washes: cameras at many sites, one zone per camera, the number of vehicles in real time and licence plate recognition at the entrance. Project experience of our team; the interface was translated for this page, sites, people and plates are anonymised.
PROJECT EXPERIENCE: SAFETY IN THE PLANTHelmet, danger zone, fire.
Helmet, danger zone, fire.
On the plant’s existing cameras.
For a processing plant, our team implemented video analytics for occupational safety on 18 cameras. The system detects people without helmets, people in danger zones and open fire, and reports every event with an image to the plant’s systems. A second AI checks the alerts once more and filters out false alarms. Plant and people are anonymised.
94.5 %Helmet detection quality (mAP50 metric)
93 %of reported helmets were correct
91 %of the helmets present were found
5.5 msComputing time per image on a server graphics card
Helmet detection measured on test images from one site (110 images with 221 helmets, February 2025) that were not used for training; values may differ in other environments. Computing time on an NVIDIA T4 graphics card, excluding pre- and post-processing, video decoding and network. Fire detection is a supplementary visual early warning, not a replacement for a fire alarm system under EN 54 or VdS. The danger-zone warning is an assistance system, not a safety function under ISO 13849.
LIVE: COUNTING PORTIONS ON VIDEOHow many portions went out today?
How many portions went out today?
The camera keeps count.
For a takeaway business, an AI counts the portions prepared in a marked zone on the existing kitchen camera. Results arrive every 30 minutes via a messenger bot and can be reconciled with the till system. No new hardware, no tally sheets.
≈ 95 %agreement of the daily total in testing: 35 counted, operator figures 33 to 37
79 %exact matches per 30-minute slot after optimisation (previously 59 %)
30 minevaluation interval, results via messenger bot
Measured at one site against the operator’s figures. This accuracy requires a correctly positioned camera above the counting zone. In Germany we agree camera analysis at the workplace with the works council.
LIVE SINCE 2026: PARCEL SHOPS & PICKUP POINTSWhat happens at the counter.
What happens at the counter.
As a report every hour.
For an operator of pickup points, the system analyses the existing cameras at several sites: opening time, pickups and their duration, courier deliveries, queues and peak times. For use in Germany we analyse processes, not individual employees, and agree the concept with the works council.
hourlyreport per site via messenger bot, plus alerts for camera outages or boxes left unpacked
42,259queue measurements and 11,845 events by mid-February 2026
under €2AI cost per camera per month in operation, because the AI analyses only one frame every two minutes
PROJECT EXPERIENCE: ACCESS & PARKING · EURecognise the plate.
Recognise the plate.
Open the barrier. Pay.
From 2017, our team developed parking terminals, barrier control and its own licence plate recognition for supermarket car parks in the EU and commissioned them on site.
- Licence plate recognitionOur own recognition at the entrance; vehicles on an approved list are let in automatically.
- Barrier & flowControl of the barrier and the order of vehicles at the entrance.
- Pay on siteParking terminals with cash and card payment, hardware and software from one team.
PROJECT EXPERIENCE: THERMAL IMAGING AT THE PERIMETERDetect people and vehicles.
Detect people and vehicles.
Even in the dark.
In 2022 our team developed tracking of people and vehicles on thermal cameras as the basis for alarms at the site boundary. People are usually not identifiable in thermal images, which makes alignment with data protection and the works council easier.
- Background modelMoving heat sources are separated from the static background, also at night and against the light.
- Kalman filterTracks are smoothed; briefly hidden people and vehicles keep being followed.
- Matching across framesThe Hungarian algorithm assigns every detection to the right object. Implemented in C++, close to the camera.
FOR MUNICIPALITIESCamera on the vehicle.
Camera events for municipalities ↗Camera on the vehicle.
Events for the town.
Fly-tipping, skips in the street, damaged signs: detected on the trips of sweepers and refuse trucks, reviewed by a person.
FREE FIRST CHECKCheck first, then decide.
Check first, then decide.
Free of charge.
Before you spend any money, we look at your workflow. You get an honest assessment and a fixed price for the next step.
Call, 30 to 60 minutes
You show us the workflow in a video call, ideally with a short phone video or a few photos in advance.
Outline within two weeks
Approach, feasibility, limits and a fixed price for the next step, a check or a pilot. Plus a note on whether a funding programme could fit.
You decide
No obligation. “Better not to do it” is a possible result too.
PROJECT CONCEPTS
Workflows you can tackle with this.
Examples with workflow, deliverable and pilot scope. They are concepts for a first pilot, not completed client projects.
Robotics & operations
Detect a Defined Visible Manufacturing Defect
We develop a camera inspection station for one component and a clearly defined defect. The pilot compares detection with your quality team's assessment and provides a documented basis for use in production.
See the workflow & deliverables ↗Robotics & operations
Check Kit Contents Before Dispatch
We set up visual inspection for a recurring shipment or assembly kit. A camera and, where needed, item identification help your team review discrepancies before packing and document the approved contents.
See the workflow & deliverables ↗FAQ
Before we start.
Can we use our existing cameras?
That depends on image quality, angle, lighting and technical access. We assess your equipment before proposing additional hardware.
What happens when people appear in the footage?
Purpose, image areas, access and retention are defined before use. Where people are affected, we work through the requirements with your responsible teams. The pilot addresses a defined operational event.
Can video analytics replace a safety system?
Ordinary video analytics is not a validated safety function. The pilot evaluates detection and alerts; requirements for a safety application need a separate assessment.
GUIDES
What customers ask before they start.
AI Video Analysis: Requirements and Process in Operations
Discover the requirements for AI video analysis in your company, how a pilot project works and what the pilot and operation cost.
NEXT
Related to this page.
Where this already runs
What it costs
Who it is for
THE NEXT STEP
What should work better in your business?
Tell us about a process that takes up your team’s time. We discuss where automation could help and what is needed to define the next step.
Describe one specific event and prepare details of your cameras, lighting and desired response. Share sample footage only after we agree on the data to be released.
Your effort: one person who knows the workflow, access to examples, a kickoff conversation and the decision after the pilot. We handle integration, setup and coordination with vendors. What we take on ↗
HEINRICH SCHNEIDER / DDC
heinrich@ddc.companyPrefer email? Write to us directly.
Contracting party: M2D GmbH, Am Turm 11, 53721 Siegburg · initial call free of charge · Pricing · Security & data protection