The Grand Google Cloud Derby Tour - Stage 1

Cloudreach Data Architect, John Heal, took the Google Cloud Derby on the road recently, heading to the British Motor Museum to run a session designed to teach participants about GCPs machine learning tools.

What’s the best way to show people the potential of Google Cloud? Clearly the answer is with robots! So, when I was approached to take the lead for Cloudreach on the first Google Cloud Derby to take place outside of a Google office, I jumped at the opportunity.

This is the approach Google and Cloudreach took, hosting a special Derby at the British Motor Museum (BMM).

Training the Cloudreach team

The Cloud Derby is a hack day created by the team at Google Cloud in America, in which the participants use a Raspberry Pi, a GoPiGo3 kit and Google Cloud Platform (GCP) to create self-driving cars. Using GCPs machine learning tools, the cars must be able to identify and capture certain coloured balls. At the end of the session, the cars are pitted against each other to decide the day’s Cloud Derby Champion!

For a crash course on what is like to participate in a Cloud Derby, Ignazio Pollina has shared his experience previously.

If we were going to successfully run this day surrounded by such iconic vehicles, we were going to need to become pros, so we spent some time with Google Cloud getting up to speed.

After reading some introductions, we met with the team at Google Cloud who we would be working with on the day. We figured the best way to learn is by doing, so we got the instructions, split into teams and began. By the end of the first day we had successfully trained the machine learning models on Tensorflow and deployed our driving applications to App Engine (Google Cloud’s managed application platform). After a bit more work on the hardware, tinkering with numbers and electrical tape, we had ourselves some automatic ball collecting robots.

We were ready to take this show on the road!

Google Cloud Derby Mini

The day of the competition arrives

Following a hearty breakfast and coffee (or three in my case), we set off to the BMM to get everything set up for the 40 participants. As the teams started to arrive, the room began to fill with curiosity as attendees first saw the robots on the tables and a huge battle arena in the middle of the room. After Google Cloud introduced the derby concept, we gave everyone a live demo of what they would be aiming to beat. Then the teams were off!

It was a quiet start as the teams began setting up their environments but they were soon starting up their own Compute Engine instance and taking loads of photos used to train their models.

Next, it was onto the model training, with many of the teams taking this opportunity to grab some lunch and popcorn while the Tensorflow codes did their work. Energised with food, everyone was deploying their controller apps, with the support from the Cloudreach team.

The room really began to liven up as the teams got round to starting up their cars, connecting them to their cloud-based apps and testing out the results of their work. A bit of configuration tuning in the app, and some engineering pit stops to swap out some failing parts later, the teams had fully functioning robots that were ready for the derby! In the first few rounds, the cars developed a bit of stage fright, so the teams gathered around their machines to make their adjustments for the finals. The room filled with cheers and cries as the robots drove and hunted out the balls, capturing their targets autonomously. The teams watched anxiously, knowing they had all done their best, all wanting to take home the pride (and prizes) of being number one.

At the final whistle, with the winners crowned, all the teams went home proud of what they had achieved throughout the day, excited with thoughts of what they could do with GCP going forward.

Google Cloud Derby

Louie taking a break from a hard day's work
Wrap up, and what’s next

Once the dust had settled on an exciting and successful day, and everyone had a well-deserved rest, the Cloudreach team got together to reflect on the event. Firstly, I thanked everyone in the Cloudreach and Google Cloud teams for all their hard work and time, all of which made this an incredible day! Then we discussed the future; the team was so full of ideas on how we can improve, augment, and extend the Google Cloud Derby in awesome ways (it was hard to keep track of them all). Clearly, it wasn’t just the guests who loved the event!

Are you excited to hear what some of these ideas are? I guess you’ll have to keep an eye out for Stage 2!

  • google-cloud-platform
  • machine-learning
  • google-cloud-derby