Facial Recognition & Interfacing with A.I.

I have an A.I. system running on a Raspberry Pi. I would like to use a Jetson Nano for facial recognition. I have the following questions:

  1. Is the Jetson Nano capable of identifying specific individual human faces?

Note: I only need it to be able to recognize about a dozen different individuals, and it would be introduced to each face for learning purposes. But I’m curious how well it could do this.

  1. Would the Jetson Nano be able to continually upgrade its recognition of these faces as it continually interacts with these people?

  2. The A.I. of my robot resides on a Raspberry Pi. All I want the Jetson Nano to do at this time is recognize specific people and be able to report the results to the Raspberry Pi.

I’m curious if the software on the Jetson Nano allows for a simple communications with the Raspberry Pi? What I envision is the Jetson Nano simply sending the name of the person it sees to the Raspberry Pi. The Raspberry Pi will take it from there in terms of any further interaction with the outside world.

In other words, all I want the Jetson Nano to do in this project is recognize people’s faces and make the results of that recognition available to another program, in this case as Raspberry Pi.


Yes, with the right software and training models. Some models that are a good start come with the system, but you will have to do significant work on your own (especially in preparing the training data for the individuals in question.)

Not really. You could build some software that looked at data from detections, and extracted new training data (especially for times of “glitches” where there is a detection before/after a frame without detection) and then augmented the corpus and then ran additional epochs of training. Training ON the jetson is slow, but it could do that in the background. You’d have to build all this yourself.

Easiest would be to use the Ethernet ports, or perhaps the built-in hardware serial ports on the devices. However, if you really wanted to, you could also write and debug more code to make I2C or SPI busses work.
I’d probably go for the serial ports, assuming I didn’t need those for anything else on the Raspberry Pi.

Sounds good. Thanks for the quick reply snarky. I don’t mind building my own code as long as I can work with the facial detection software that comes with the Nano. I don’t yet have a Jetson Nano so right now I’m just looking at what the potential is. I definitely want a system I can expand on.