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.
Comments