Skip to content
← Back to cinfo.es
Catalogue

SERVICE · ARTIFICIAL INTELLIGENCE

Computer vision

We turn any camera into a sensor that understands what it sees.

Counting, detection, classification and alerting over live video, in the cloud or on the device itself.

IndustryRetail

Contact cinfo about Computer vision Download data sheet (PDF)

We take you to cinfo's contact form with this item already filled in, so you don't have to explain what it is about.

Computer vision

THE CHALLENGE

Cameras record plenty, and almost nobody watches

A site full of cameras only helps you look back after something has already happened. Turning that image into useful information in the moment means analysing it automatically.

  • Too many hours of video

    No team can monitor dozens of cameras continuously and reliably.

  • Data nobody collects

    Occupancy, flows, frequencies and patterns that are in the image but never counted.

  • Late reaction

    Incidents are found when reviewing the recording, when there is no time left to act.

WHAT IT IS

AI applied to any video input

cinfo builds computer vision solutions on the cameras a client already has or installs: the models detect objects, people and vehicles, generate statistics and raise alarms, in the cloud or directly on the device.

Models trained for the specific use case, not generic off-the-shelf ones.

AI deployable at the edge when latency or bandwidth demand it.

A complete dashboard with reporting, alerts and historical statistics.

Integration with sensors and with signalling or access control systems.

AI applied to any video input

HOW IT WORKS

From use case to a model in production

Every project starts from one concrete, measurable use case, not from a technology in search of a problem.

  • We define the case

    What must be detected, with what accuracy, and what should happen when it is.

  • We assess the cameras

    Position, angle, resolution and lighting determine what can actually be detected.

  • We train and validate

    Models tuned to the real setting, validated on footage from the site itself.

  • We deploy and measure

    Go-live in the cloud or at the edge, with dashboard and alarms.

SOLVED CASES

What we have already put into production

  • Loading docks

    Counting boxes loaded and unloaded, detecting forgotten packages and alerting the warehouse.

  • Abandoned objects

    Detection and recording of objects left out of place, with AI at the edge.

  • Parking control

    Occupancy, free spaces, double parking, dwell time and vehicles in restricted zones.

  • Ports and moorings

    Mooring time control and detection of boats in unauthorised areas.

  • Pedestrian crossings

    Pedestrian detection integrated with audible and light signals, and police alerts.

  • Smart bus shelters

    People counting by time slot, boardings and alightings, and service frequency.

  • Traffic modelling

    Vehicle classification, direction of travel, patterns, anomalies and plate reading.

  • Sports analysis

    Trajectories, positioning, technical gestures and player identification.

BENEFITS

The camera stops being an archive and becomes a sensor

  • Immediate reaction

    Alerts the moment the incident happens, not when reviewing the recording.

  • Data where there was none

    Counts and patterns that let you size services and justify decisions.

  • It reuses the installation

    In many cases the cameras the client already has are enough.

  • Our own technology

    Models developed in Spain, with data hosted in our own datacentre.

SCOPE

What a computer vision project includes

A project can be limited to a proof of concept on one camera, or cover full deployment with dashboard and alarms.

  • Feasibility study

    An analysis of the available cameras and what can be reliably detected.

  • Proof of concept

    A model trained and validated on real footage from the site.

  • Production model

    Deployment in the cloud or on an edge device, at the agreed performance.

  • Dashboard

    Visualisation of counts, events and statistics, with report export.

  • Alarm system

    Configurable rules alerting the relevant people or external systems.

  • Integrations

    Signalling, access control, sensors and the client's management platforms.

METHODOLOGY

How we approach a new case

  • Feasibility

    A review of the installation and definition of measurable success metrics.

  • Training

    Model tuning with footage from the client's real setting.

  • Validation

    Results checked against manual observation before detection is accepted.

  • Production

    Deployment, performance monitoring and retraining if the setting changes.

Achievable accuracy depends on camera placement and angle, resolution and lighting. That is why the feasibility study always comes first.

GALLERY

Detections in real settings

Counting and control in loading docks and warehouses.
Counting and control in loading docks and warehouses.
Vehicle detection and occupancy control in car parks.
Vehicle detection and occupancy control in car parks.
People counting at public transport stops.
People counting at public transport stops.

CASE STUDIES

Video AI projects in production

  • SPANISH VOLLEYBALL FEDERATION

    Tactical analysis of beach volleyball

    Analysis of ball trajectories and player movement and positioning, with each technical gesture coded and rated, plus team and player statistics.

    SPANISH VOLLEYBALL FEDERATION

  • PORTUGUESE FOOTBALL FEDERATION

    A player catalogue by match

    A player identification proof of concept: a player is tagged in one video and the system recognises them in the federation's later footage.

    PORTUGUESE FOOTBALL FEDERATION

  • COUNCILS AND URBAN OPERATORS

    Traffic, bus stops and pedestrian crossings

    Vehicle counting and classification, traffic flow modelling, people counting at stops and pedestrian detection with audible and light warnings to vehicles.

    COUNCILS AND URBAN OPERATORS

CINFO IN NUMBERS

Analysis capability

  • 0,1 svideo AI reaction time
  • Edgeon-device deployment when required
  • 39NVIDIA GPUs available for training
  • 12weeks to calibrate a new scenario

FREQUENTLY ASKED QUESTIONS

Questions before starting an AI project

Will my current cameras do?

Often yes, but it depends on placement, angle, resolution and lighting. The feasibility study is always the first step.

Does the video have to go to the cloud?

Not necessarily. When latency or bandwidth demand it, the AI is deployed directly on an edge device next to the camera.

What accuracy can I expect?

It is set as a success metric at the start of the project and validated against manual observation before detection is accepted.

How long does a new use case take?

It depends on the scenario. As a reference, calibrating the sports tracking AI for a new discipline takes around twelve weeks.

What about data protection?

Processing is designed in line with applicable video surveillance and data protection regulations, and data is hosted in cinfo's own datacentre in Galicia.

Does it integrate with my systems?

Yes. Detections can trigger audible and light signals, open barriers, alert an owner or feed the client's management platform.

WHO IT IS FOR

Where we are deploying video AI

  • Cities and mobility

    Traffic, public transport, pedestrian crossings and low-emission zones.

  • Logistics and industry

    Loading docks, warehouses, safety and object control.

  • Ports and car parks

    Occupancy, dwell times and restricted area control.

  • Sport

    Tactical analysis, sports performance and player identification.

Related items

LET'S TALK

Let's start with one camera and one concrete case

Send us some footage from your site and we will tell you what is detectable, and how accurately.

Request a feasibility study