Programmer completes cycle journey of UK in virtual reality with Google data
Tue 24 Jan 2017
A software developer who created a ‘Virtual Britain’ using Google Street View data has completed a 932-mile journey across the country from the relative comfort of his home.
Aaron Puzey used a Gear VR headset in combination with a low-cost Bluetooth cadence monitor and an Android smartphone to generate a virtual cycle route down the traditional run from Land’s End to John O’Groats, whilst apparently passing through a 3D representation of the landscape and streets, as derived from the Google data.
Puzey began the trip after work in May of last year, putting in 85 hours at an average speed of 10.9 mph, burning more than 50,000 calories and passing through over 300 virtual villages over 22 counties. Milestones are marked by floating boxes.
The route is set in advance, or ‘baked’, allowing for limited movement within a sequence of distinct VR spheres, but not allowing the user to digress from the planned path. Puzey is inviting suggestions at the project’s site for new potential routes around the world and plans to bring the capability for more ad hoc routes to the initiative.
Puzey notes that the cyclist needs to avoid deviating much from the centre of the illusion in order to avoid seeing trees or other distinct promontories that are actually plastered onto walls and planes – since they do not really contain 3D data.
The software uses the Unity game engine to pull in and assemble 8 data segments which form the current sphere for the rider, with the source data having been converted from GPS coordinates to meters. Initially this caused a hiccup in transition while Unity assembled the segments, nearly causing the programmer to abandon Unity for the Unreal game engine – but finally the problem was solved with preloading.
Puzey solved some of CycleVR’s thorniest initial challenges with work based on Paul Wagener’s Streetview Explorer code, dedicated to creating full-transition VR spheres from static data, with often bizarre results:
What’s interesting about Puzey’s use of the data is how smooth the transitions are between the stop-frame style source data, with the routes – complete with the customary blurred faces and car number plates – appearing as video footage instead of the ‘slit-scan’ transitions the rest of us see when navigating Street View. Google has conducted its own research into ‘in-betweening’ Street View images, which are frames actually taken from Immersive cameras mounted on Google vehicles, selected at GPS intervals of approximately four metres.
Since streaming the source video – besides mitigating against an interactive experience – would challenge even Netflix in terms of bandwidth demand, it’s ironic that even Google itself has to reinvent video from images taken from its own video.