Google Earth VR vs Open Street Map VR
By Gabriel Rizk
Last summer, I worked at a live courier service called Quiqup, where I built a neural network that chose which mode of transportation (MOT) was best to fulfill a given order (small orders required scooters/bikes, large orders required cars.) The largest issue was that the orders came in as mostly textual data, so I had to use a number of NLP techniques to classify the orders among the delivery types. I also had a lot of metadata to use, such as the pickup origin, and I wanted to see how much bearing this metadata had on the chosen mode of transport. I wanted to put a map overlay on Google Earth VR to be able to better visualize the pickup and dropoff locations to see if the odd design of London streets would have an effect on the outputs.
Google Earth VR currently does not have a way to put a map overlay on top of its existing map. I tried to find numerous workarounds for this, but nothing seemed to work because Google did not provide a usable way for this to be done.
Open Street Map does have a way to put an overlay over its own maps, but it does not look pretty. I have attached a screenshot below of how it works, as well as a video showing how to put Open Street Map data into VR using A-Frame
Results of KNN Clustering of Locations
• Certain stores and locations clustered with specific MOTs
• Takeout food and groceries had smaller MOTs
• Clothing (laundry, stores) tended to have larger MOTs
• Assorted goods had a lot of variance, quantity was the most impactful factor in clusters
• Biggests Clusters: Grocery stores, alcohol stores, malls
• I initially wanted to do it in Google Earth VR, but there is actually no way of doing it as of now.
• I attempted so many ways to try to work around this, and attempted to overlay the data in Google Earth, but the functionality didn’t exist and was not implementable (lost many hours researching and attempting this)
• I eventually settled on Open Street Map, which is not nearly as impressive (image above)
• Project changed from a visualization of the clusters to a review of the software.
Google Earth VR
Visuals are extremely beautiful, and provide extensive information about a city, as if you were walking in its streets.
Most of the world is modelled. Every major and most minor cities are modelled, only inaccessible and/or small nations are still poorly covered (i.e. China, North Korea, small African nations)
Gaping use case hole: No ability for individuals to overlay their own data on Google Earth's map. This would be game changing, as companies that require nuanced knowledge of a city (Uber, Lyft, etc) can gain this without having to spend the resources to be on the ground.
Open Street Map VR
Open Street Map in VR fulfills the gaping hole that Google Earth VR desperately needs to fill, as no other company has essentially mapped most of the world as Google has with Google Street View.
You are able to put in your own data, and visualize it in VR, but the results are no where near as pretty or useful as Google Earth is (visualization below)
You do gain information with regards to the input data, as you can see the direction of roads, surrounding structures, and such, but you lack the nuance and fine grained detail that Google Earth VR has. You also can not tell, for example, how many lanes this road contains, and other pieces of information that can be used by people looking to navigate a city efficiently.
• By working on this project, I found this gaping hole in a major use case for VR
• Google Earth VR has massive potential if it can fill this use case.
• Open Street Map is very niche, but fulfills these use cases, just not very well.
• Geospatial data overlays can have major uses for companies that require nuanced knowledge of a city
• Uber to know what streets are most efficient, Quiqup to optimize delivery times and MOT choices, etc.