Object detection in camera footage, from pilot to production.

We train models to find objects in images and video, using the cameras you already have or new ones. The result becomes an alert, a count, or a record in your system.

Airport baggage carousel with suitcases and backpacks marked by detection boxes
  • Camera 01suitcase
  • Camera 02backpack
  • Camera 03suitcase
  • Camera 04backpack

Simulated example

Object detection projects

For confidentiality, projects are shown without client names.

Airport baggage carousel with suitcases and backpacks marked by detection boxes
Simulated image

Suitcases and backpacks in airports

Identifying luggage in camera footage.

Car at a gate barrier with its license plate marked by a detection box
Simulated image

License plate recognition

Reading the plate in the camera image.

Students seen from behind in a classroom, with a phone on a desk marked by a detection box
Simulated image

Phones in classrooms

Detecting the device in classroom images. The focus is on the object.

LibrasLink: computer vision on the device itself

Libras in real time

An app that translates Brazilian Sign Language (Libras) into written Portuguese using a phone camera.

Manual alphabet on the device

In the MVP, the model recognized the Brazilian Sign Language alphabet running on the phone.

Images never leave the phone

Only the key points of the gestures are processed off the device. Top 10 at Startup One FIAP.

How we develop it

  1. Define what needs to be seen

    The object, the environment, the camera, and what should happen when it appears.

  2. Collect and label images

    We use images from your real environment, with the actual lighting and angles.

  3. Pilot with acceptance criteria

    We measure correct detections and false alarms in the field, against a target agreed up front.

  4. Integration

    The result reaches your system as an alert, a count, or a record.

  5. Ongoing monitoring

    We monitor quality and retrain the model when the environment changes.

Cameras demand care with data

Legal basis before the project

We define the purpose and legal basis with you under the privacy law that applies to your business.

Keep only what's necessary

Retention periods, access control, and usage logs are defined in the project.

Where the model runs

On the device, on a local server, or in the cloud, depending on risk and volume.

Working with us from outside Brazil

Tailored contract

The proposal sets out who signs, the governing law, and the jurisdiction.

Invoicing and currency

We agree on the currency and how you are invoiced.

Data

We sign a data processing agreement when needed and choose where your data lives.

Hours

Brazil overlaps with US business hours, and we set meeting and support windows with you.

Questions

Do I need to buy new cameras?

Not always. We assess the footage you already have and tell you if the quality is good enough for what needs to be detected.

How do I know it will work in my case?

We start with a pilot in your environment, with acceptance criteria defined up front and measurement of correct detections and false alarms.

How much does it cost?

Every project is quoted individually. The cost depends on what needs to be detected, the volume of images, where the model runs, and the integrations.

What should your cameras be seeing?

In 30 minutes, we'll assess whether the project is viable and how to test it.