This repository houses the REF RaAS re-implementation of the research software for the host portion of the Coral Spawn & Larvae Imaging Camera System (CSLICS) found here.
This is also inclusive of the DepthAI pipelines for running on the Luxonis OAK series cameras as the host is responsible for deployment.
Install the cslics_mqtt package.
The following is the proposed system architecture visualised with draw.io from this file.
The following describes the communications between programs in the CSLICS system. Visualised with draw.io from this file
The following is the database schema which has been visualised with PlantUML from this file.
The following is the proposed architecture for the Science Mode operation visualised with draw.io from this file.
I recommend using Mosquitto as your MQTT broker. A package for it is provided in the Ubuntu apt repositories.
# apt install mosquitto
As we will be using this on a closed network, we won't be interested in any security on the communications.
Create a configuration file containing the following:
listener 1883
allow_anonymous true
To run the broker, run the following:
$ mosquitto -c /path/to/mosquitto.conf
As per the documentation on the Luxonis website, models must be first converted to a MyriadX blob after an optimisation stage through OpenVINO IR.
As OpenVINO has in more recent version removed some of the necessary tools from their packages, this is most easily done (locally) by running up the provided image with Docker Compose: https://github.com/luxonis/tools
Ensure all steps in the README.md are followed!
This was last tested to work with a YOLOv8 model on this commit
Note: Only detection models are supported for YOLOv8!