Your robot can tell who you are
A new face_recognition card enrolls a face from a photo or a link and then says who is in frame, with a confidence figure. Inference and the face database both stay on the robot; JetPack 5.11 machines run it on the GPU.
Your robot can now tell who is standing in front of it. The perception bundle gains a face_recognition card that ties a face to a name, so the next time it sees you it can greet you by name.
Using it
Drop face_recognition on the canvas and wire it to a camera output.
| Action | What it does |
|---|---|
register_by_photo | Enroll from a local photo — the card renders a file picker |
register_by_url | Enroll from an image link |
recognize | Say who is in frame, and how confident it is |
You can also just ask for it in conversation: tell the robot to find a picture and enroll it. That path works end to end now — a local file the agent produces is uploaded to the perception service automatically, with no manual upload step.
It runs on the robot
Inference happens on the device and the face database stays there. No face ever leaves the machine.
The card's device setting defaults to auto: JetPack 5.11 machines run it on the GPU, JetPack 6.1 falls back to CPU for now. Either way recognition works — the only difference is speed.