Skip to content
SwiftSCIENCE

Custom behavior AI

Track the behaviors your study needs.

SwiftAi builds a model for the behavior you measure, from your own footage. People review what it proposes.

A rack's side camera: a detection model looks at every frame, boxes each mouse and follows it, frame by frame.

What a model sees

Found and followed, by day and with the lights off

With the lights off, on an infrared camera: three mice in a cage, each boxed, numbered and followed.
An earlier behavior classifier on archived footage from a research study, shared with permission. It proposes a behavior for each moment; the number is the model's score for that proposal, not an accuracy. People check every proposal against the footage.

From footage to reviewed results

  1. 1

    Your footage

    From SwiftSENSE cameras, or the video your lab already has.

  2. 2

    A model for your behavior

    Trained for your cages, your camera angles and the behavior you define.

  3. 3

    People review

    Proposals are checked before they count, then exported with the study's groups.

For science

Custom behavior tracking for your endpoints, delivered as reviewed data, scoped algorithms or a final report.

For care teams

The same cameras let people check every cage on video from outside the room (SwiftSENSE). A model never stands in for that check.

Questions

Can we use SwiftAi without SwiftSENSE?

Yes. We can analyze footage your lab already has, or capture it with our cameras.

Who checks what the model finds?

People. A model proposes; your team or ours reviews before anything counts.

Does it run in the cloud?

On SwiftSENSE Edge, models run at your site. Cloud AI is a separate choice you make.

How do we start?

Tell us the behavior you need to measure and what makes scoring it difficult. Our research services can run the whole study, or only the stage you need.

Which behavior do you need to measure?

Tell us the behavior and what makes scoring it difficult. We will use your answer to choose what to show you.