2-face-detection

Lab 2: Face Detection

We will now deploy our first machine learning model to the device using Greengrass.

Switch to Lab 2 directory

If at the root of the repository, run the following commands to switch to the current path for the lab files by running the following in your shell:

cd GG-Edge-Inference
cd 2-face-detection

Also, if you are running this lab on Cloud9, make sure that your PATH is correct. You can do so by running the following command -- this is not necessary or even recommended if you are running on your own local machine:

export PATH=/opt/c9/python3/bin:$PATH

Make

1. At the root of the workshop repository path -- ie. the parent directory of this lab -- there is a file called properties.sample.mk. We will be copying this sample file for our lab by running the following command in your shell (assuming that you are in the current lab path):

cp ../properties{.sample,}.mk

1. The newly created file, called properties.mk, will have specific sections that you will need to edit to match your own settings:

   • REGION: should match your AWS region (default is us-east-1)
   • BUCKET: should match the Amazon S3 bucket you created as part of the workshop setup
   • GG_GROUP: should match the name of the AWS Greengrass group you created during Lab 1 (default is ml-edge-workshop)
   • LAMBDA_FUNCTION: should match the name of the AWS Lambda function you created during Lab 1 (default is ml-edge-workshop-lab-1)

2. Once you have edited the file to match your own configuration requirements, we can now run the make command in the current path. This will execute a make operation against the Makefile in the current path, which references the properties.mk file in the parent path.

3. Go in the first directory 2-face-detection.

4. Run make. This will package, then deploy, the new Lambda function and all the required libraries to perform face detection on your device.

See the results

1. In the IOT console, got to the "Test" section.

2. Subscribe to the topic face_recognition/inference, it will show you who the device sees. There are more topics available:

   • face_recognition/inference: the faces it sees, NewX where X is an increment.
   • face_recognition/admin: some info messages or errors if there is.
   • face_recognition/new: where the device posts faces it doesn't know (Base64 encoded jpeg).
   • face_recognition/match/DEVICE_NAME: used to update the name of a face in a device.

   You can also use the symbol # to use as a wildcard for example face_recognition/#.

3. To rename a user: post a message to the face_recognition/match/DEVICE_NAME with content {"OLD_NAME": "NEW_NAME"} where for example OLD_NAME would be "New0" and NEW_NAME is your name.

Bonus L33T Trick (Local Users with Linux/Mac and mplayer installed)

1. After the deployment is done you can view the output with: ssh <DEVICE-IP> cat /tmp/results.mjpeg | mplayer - -demuxer lavf -lavfdopts format=mjpeg:probesize=32

There is currently an issue with this function that makes it freeze for around 30 seconds the first time it sees a face, however this only happens once.

You should be able to see the results of the inference both in the topic and the video stream. Great! Now you've used the basic components of Greengrass, let's move on and start doing your own models.