Crowdsourced Maps Could Be Useful For Self-Driving Cars

Google Maps might be helpful when you want to find a restaurant or shopping mall. However, if you decide to have an autonomous car driving you to your dining place or your favorite shop for a shopping spree, it is going to require a much more accurate map.

A startup with a funny name, Mapillary, thinks that it’s discovered the best way to create that map.

Mapillary says that crowdsourcing combines with artificial intelligence can generate highly accurate and constantly updated maps of roadways, complete with other details like driveways and road signs, the exact kind of maps needed for self-driving vehicles.

Contributors to Mapillary can upload photos of driveways, streets and other objects that an autonomous vehicle might encounter. A computer-vision software of the company then makes an analysis of the photos. Moreover, it needs to take into consideration every other aspect like manhole covers to utility poles.

A licensing fee is required if commercial customers want to access these highly accurate and detailed maps of Mapillary. However, the maps are free if they are used for charitable, educational or personal purposes.

When being asked how to keep people contributed to the maps, the company replied that periodic mapping challenges are held frequently with prizes for winners. There is also a leaderboard for people who have made the most contribution.

Mapillary was founded in 2013. Until now, 422 million images from 190 countries have been collected, generating maps of over 3.9 million miles of roadways. Unlike Google Maps, Mapillary Maps is updated constantly, therefore, it is suitable for self-driving cars.

In an interview with MIT Technology Review, Mapillary’s CEO, Jan Erik Solem said: