Northern Ireland's First Space Data Hackathon
Date posted
5 March 2018
Reading time
12 Minutes
Northern Ireland's First Space Data Hackathon
Space Data Hackathon
Over the weekend past (23-25 Feb), Code4Good NI in collaboration with Kainos hosted a hackathon at the new computer science building at Queen's University Belfast. The hackathon was themed on Earth Observation data, inspired by one of the key themes of the Northern Ireland Science Festival: Space. Why is Kainos interested in space anyway? Well it seems pretty obvious when you consider how much they have invested in artificial intelligence and AI Camp. These days, there are a lot of notable Earth Observation data providers out there such as Airbus and Copernicus.
*Jk, of course I know Stormont!
What is Earth Observation data?
EO data is gathered by satellites orbiting Earth. The satellites use optical imagery as well as radar sensors to get extremely high resolution 3D imagery of the planet. Airbus kindly provided some optical and radar images of Belfast, Enniskillen, and Derry for the hackathon. However, EO data is notorious for being difficult to work with. Some of the challenges that exist are:
The Team[/caption]
The Kick-Off on Friday saw a presentation by Deirdre Ferguson from Smart Cities Belfast to give some hints on the judging criteria and inspire some ideas by proposing problems that needed solved. Well, actually it was a presentation on all the brilliant work that Smart Cities have been doing for Belfast and if you've been to a hackathon before, you know these are the things you have to keep note of if you want to impress the judges!
Lastly Luke McNeice, the Innovation Lead at Kainos, gave a presentation with some mind blowing facts about EO data to inspire and some very useful hackathon tips, before setting us off to form teams.
I had brought a couple of my friends along but we were keen to expand our team so we created a 'WANTED' poster to attract new team mates and left it overnight.
Returning on Saturday, we gained 3 new team mates. Success!
The six of us grabbed a table in the neat garden room on the second floor of the computer science building (there's a whole wall of moss in this room!) and began brainstorming.
We aimed to:
Screenshot of the vision processing tool tagging the bonfires and nonfires[/caption]
The team banded together and, using QGIS, we scanned the satellite imagery for bonfire sites in Belfast and labelled them as 'Bonfire' in the vision processing tool, gathering plenty of 'Nonfire' images of the city as well.
After gathering nearly a hundred images we were able to get around 95% accuracy finding bonfires in an image with the vision processing tool. Pretty sweet result!
We were feeling confident with our pitch on Sunday, but the competition was tough! There were around 10 teams who presented their ideas and while we didn't win first prize, I enjoyed the weekend and it was a great opportunity to work with great software engineers from all over.
The winning team was Team Glaziers proposed the use of satellite data to get more accurate locations of potholes in Northern Ireland to speed up the process of locating and fixing problems in the roads. A solution we can all get behind I'm sure!
Summary
Big congrats to Team Glaziers for their winning idea!
A few things I learned this weekend:
Team Selfie![/caption]
- the tools required to process the data are hard to use
- the development environment is difficult to set up
- the sheer size of data**

- Solve a problem for Smart Cities Belfast
- Build a useful dataset
- Improve council asset management
- Make city living better for residents
- Knowing all the bonfire locations
- Measuring proximity to residential areas
- Detecting illegal materials, such as tyres
- Detecting a bonfire that's getting too large earlier


- Space data is indeed hard. The images we were given were so huge that sometimes they'd crash our computers and the only tool we could use to display them was QGIS - which is very intimidating if you've never used it before! Thankfully, the teams had Jordan McDonald for support with using the tools.
- Preparation of space data is difficult and it's important to have a large enough and clearly defined dataset for machine learning to really work. There was a lot of trial and error in our team while trying to create a dataset of bonfire and nonfire images - probably about 95% of our effort was spent on this!
- The Belfast Tech Community have a great collaborative spirit! Despite competing, everyone helped each other out troubleshooting problems with tools or data. It was great to be a part of it!

Thanks to my team for a fantastic weekend and all the hard work:
Mark Bailie (Core Systems NI)
Robert Beck (Student)
Claire Burn (Rapid7)
Zoe Gadon-Thompson (Student)
Rebecca Martin (Student)
And special thanks to Jordan McDonald for supporting so many teams and so well!