Swedish Mapillary is opening a new research lab in Austria, which will focus on advanced computer vision to enhance image recognition technology of the startup which is building a rival to Google’s Streetview, based on crowd-sourced photos.

The new lab will be headed by Peter Kontschieder, a Microsoft Research Cambridge veteran who last year was awarded the prestigious Marr Prize in machine vision for proving a new technique for teaching computers to recognise images.

Mapillary taps into local communities, crowdsourcing photos from smartphones or action cameras and open sourcing its tools – creating a platform for crowdsourced virtual reality.

Since launching in late 2013, Mapillary has mapped a million miles of roads with 61 million pictures. Its machine vision technology stitches those millions of images together to create immersive ground­-level maps and constructs locations in 3D.

mapillary_ios_map

Mapillary joins Amazon and other firms opening research labs in Graz, a small city in Austria that has become a flourishing hub for computer vision research. “Graz has a thriving cluster of artificial intelligence research and talent, so it was a natural choice for us,” said Jan Erik Solem, CEO of Mapillary.

“We are excited to have Peter lead our deep learning team, and we will grow the team there to further improve our ability to extract useful data from images.”

In March Mapillary raised its $8 million A Round from Sequoia, Atomico, LDV Capital and PlayFair, aiming to surpass Google’s coverage in the next three years. There is some room to go – in the first five years of building Street View (2007-2012) Google mapping cars drove more than 5 million unique miles of road.

Mapillary’s founder, machine vision expert and Lund University professor Jan Erik Solem sold his face recognition software firm Polar Rose to Apple in 2013.