Skip to Content

Putting Crowdsourcing on the Map

Mapillary is trying to build a community-generated version of Google Street View.
February 28, 2014

Even in San Francisco, where Google’s roving Street View cars have mapped nearly every paved surface, there are still places that have remained untouched, such as the flights of stairs that serve as pathways between streets in some of the city’s hilliest neighborhoods.

It’s these places that a startup called Mapillary is focusing on. Cofounders Jan Erik Solem and Johan Gyllenspetz are attempting to build an open, crowdsourced, photographic map that lets smartphone users log all sorts of places, creating a richer view of the world than what is offered by Street View and other street-level mapping services. If contributors provide images often, that view could be more representative of how things look right now.

Google itself is no stranger to the benefits of crowdsourced map content: it paid $966 million last year for traffic and navigation app Waze, whose users contribute data. Google also lets people augment Street View content with their own images. But Solem and Gyllenspetz think there’s still plenty of room for Mapillary, which they say can be used for everything from tracking a nature hike to offering more up-to-date images to house hunters and Airbnb users.

Solem and Gyllenspetz have only been working on the project for four months; they released an iPhone app in November, and an Android app in January. So far, there are just a few hundred users who have shared about 100,000 photos on the service. While it’s free for anyone to use, the startup plans to eventually make money by licensing the data its users generate to companies. 

With the app, a user can choose to collect images by walking, biking, or driving. Once you press a virtual shutter button within the app, it takes a photo every two seconds, until you press the button again. You can then upload the images to Mapillary’s service via Wi-Fi, where each photo’s location is noted through its GPS tag. Computer-vision software compares each photo with others that are within a radius of about 100 meters, searching for matching image features so it can find the geometric relationship between the photos. It then places those images properly on the map, and stitches them all together. When new images come in of an area that has already been mapped, Mapillary will add them to its database, too.

It can take less than 30 seconds for the images to show up on the Web-based map, but several minutes for the images to be fully processed. As with Google’s Street View photos, image-recognition software blurs out faces and license plate numbers.

Users can edit Mapillary’s map by moving around the icons that correspond to images—to fix a misplaced image, for instance. Eventually, users will also be able to add comments and tags.

So far, Mapillary’s map is quite sparse. But the few hundred users trying out Mapillary include some map providers in Europe, and the 100,000 or so images to the service ranging from a bike path on Venice Beach in California to a snow-covered ski slope in Sweden.

Street-level images can be viewed on the Web or through Mapillary’s smartphone apps (though the apps just pull up the Web page within the app). Blue lines and colored tags indicate where users have added photos to the map; you can zoom in to see them at the street level.

Navigating through photos is still quite rudimentary; you can tap or click to move from one image to the next with onscreen arrows, depending on the direction you want to explore.

Beyond technical and design challenges, the biggest issue Mapillary faces is convincing a large enough number of users to build up its store of images so that others will start using it and contributing as well, and then ensuring that these users keep coming back.

Crowdsourced services like Waze and Wikipedia have managed to do this successfully, but as Wikipedia’s declining number of editors shows (see “The Decline of Wikipedia”), it’s not easy to keep this up over time.

Mapillary hopes the growing number of smartphone users will help, though. “I think the timing is right, both in the maturity of these devices but also from connectivity around the world,” Solem says.

Keep Reading

Most Popular

Large language models can do jaw-dropping things. But nobody knows exactly why.

And that's a problem. Figuring it out is one of the biggest scientific puzzles of our time and a crucial step towards controlling more powerful future models.

OpenAI teases an amazing new generative video model called Sora

The firm is sharing Sora with a small group of safety testers but the rest of us will have to wait to learn more.

Google’s Gemini is now in everything. Here’s how you can try it out.

Gmail, Docs, and more will now come with Gemini baked in. But Europeans will have to wait before they can download the app.

This baby with a head camera helped teach an AI how kids learn language

A neural network trained on the experiences of a single young child managed to learn one of the core components of language: how to match words to the objects they represent.

Stay connected

Illustration by Rose Wong

Get the latest updates from
MIT Technology Review

Discover special offers, top stories, upcoming events, and more.

Thank you for submitting your email!

Explore more newsletters

It looks like something went wrong.

We’re having trouble saving your preferences. Try refreshing this page and updating them one more time. If you continue to get this message, reach out to us at customer-service@technologyreview.com with a list of newsletters you’d like to receive.