alexdauenhauer/w251_hw3

face detection assignment

w251_hw3

Project Summary

This project asked us to build a IoT application pipeline using a Jetson TX2 as the "edge" running through an MQTT broker-client pipeline to store captured images in the cloud.

In my MQTT pipeline I used QoS 0. I made this decision because it is the most minimal QoS level and part of the goal of this assignement was to remain lightweight. Since I had end-to-end control over the pipeline and there was no repurcussions if data was lost, QoS level 0 seemed appropriate since it will send a message once only and do nothing to guarantee receipt. I used a single MQTT topic and named it "faces" because it seemed like it fit with the theme of the assignment.

The faces can be located here: http://s3.us.cloud-object-storage.appdomain.cloud\alex-hw3-bucket

a sample image capture is shown below

Here is a screencap of the data being saved in the bucket

To make this all work

Starting on the jetson side...

Build the dockerfiles you want. I made a dockerfile for the face detection, the MQTT broker and the MQTT forwarder

cd /path/to/Dockerfile
docker build -t myimage .

Once the dockerfiles are built, create a network to tie the broker and forwarder

docker network create --driver bridge hw03

Start up your broker. In my dockerfile I already downloaded the mosquitto packages so I only need to start up the broker

docker run --name mosquitto --network hw03 -p 1883:1883 -ti broker sh
# in the container now, spin up mosquitto broker
/usr/sbin/mosquitto

Leave this running, open a new terminal and spin up the forwarder container. Once inside, I run the forwardMessage.py script

docker run --name forwarder --network hw03 -ti forwarder sh
# inside the container, run the script
python3 forwardMessage.py

Now the broker and forwarder are listening and ready to receive messages and sent them to the cloud. Moving over to the cloud...

In the cloud instance...

ssh to the cloud instance

ssh root@ip_of_vsi

build docker containers. I used the same broker container in the cloud as I did on the Jetson, I made a new one for the processor

cd /path/to/Dockerfile
docker build -t myimage .

build a docker network to tie the broker and processor together. Start up the broker, same way we did on the jetson

docker network create --driver bridge hw03
docker run --name mosquitto --network hw03 -p 1883:1883 -ti alpine sh
# now in the container run mosquitto
/usr/sbin/mosquitto

Leave this running, open a new terminal and spin up the processor container. Once inside, I run the saveToBucket.py script

docker run --name forwarder --network hw03 -ti processor sh
# now inside the container
python3 saveToBucket.py

Everything is now ready to extract faces. Back to the jetson

Back in the jetson...

open up a new terminal and spin up the face dectection container. NOTE: need to run xhost + to allow the container to see the webcam device

xhost +
docker run -e DISPLAY=$DISPLAY --net=host --privileged -ti faces

Pick up the camera and point it at your face. Go to bucket and see that face images have magically been stored there