Jetson AI Facial Recognition

Step 4B Optional: If you are using Jetson (disregard this if you are using Intel Up2 board with Movidius)

This step requires we getting ready for TX1, upon loading up the TX1 or TX2 we can follow the following guide to install caffe as well as opencv:

https://jkjung-avt.github.io/opencv3-on-tx2/https://jkjung-avt.github.io/caffe-on-tx2/

Once the pre-requisite is finished, it is fairly easy to copy the following files from the training server to the device.

deepID_deploy.prototxt from the source code (originally from https://github.com/hqli/face_recognition), change num_output: 2 or whichever amount of face you are using.

mean.binaryproto from previous step.

snapshot_iter_2000.caffemodel there should be a snapshot every couple of hundred iterations, we will just use 2000 as our base, this will be in the folder that where you trained your model.

label.txt create a file, making first line unknown and second line you.

Once these 4 are loaded you can download the project file https://github.com/Nyceane/ai-face-lock-jetson and change them into your own files. Run following command

python3 face_lock.py

You should be able to detect your own face through this. The video below shows when facelock working in real time.