Return to the white paper

This folder contains files for setting up a group of containers on a Jetson. It is intended to replicate an edge device that estimates a latitude and longitude based on a camera's picture of the sky. In our case it we are using random synthetic images rather than an actual camera.

We set up two containers.

1. A camera container (send_image) that takes images generated using the functions of test_images and publishes them to a MQTT broker.
2. A MQTT broker (mosq-broker) that receives messages and passes them along.

A third container (inference) is described in the README at \inference. It subscribes to the mosq-broker, which means it is sent images from the send_image container.

In simulation mode, the camera container send a new image to the broker (and thus on to the inference container upon a key click. The esc button is used to exit the send_image program.

Link to a demo video: https://zoom.us/rec/play/vcEvdr39r2g3SNPDtASDV_UtW421e_-s1SUf-PYOnU-zACRQMVqmMOZGNLBaIGprp4ITygeePJec76GT

docker network create project

docker build -t broker-image -f Dockerfile.broker .

docker run --name mosq-broker --rm -p 1883:1883 -v /w251-project/inference:/inference --network project -ti broker-image mosquitto -v

docker build -t send-image -f Dockerfile.send .

Before running container: xhost + (provides access to display)

From within edge_network folder docker run -e DISPLAY=$DISPLAY --privileged --name send_image -v /w251-project/inference:/inference --rm -v /tmp/.X11-unix:/tmp/.X11-unix --network project -ti send-image bash

From within container bash image.sh