Computer Vision with Peter Kontschieder

1:01:30
 
Share
 

Manage episode 220733073 series 1437556
By Discovered by Player FM and our community — copyright is owned by the publisher, not Player FM, and audio streamed directly from their servers.

Mapillary is a company that processes high volumes of images to develop a labeled 3-D model of the physical world. Mapillary’s APIs allow developers to build applications that are aware of stop signs, buildings, streets, trees, and other physical objects in real-world space. The potential use cases for Mapillary are numerous, ranging from self-driving cars to augmented reality.

We can now build a 3-D model of the real world. It’s not a perfect representation of reality, but it is much better than we had just a few years ago. What has changed? How have the tools advanced such that we are able to build an API for accessing accurate information about the physical world around us?

Mapillary is possible because of a combination of modern developments.

High quality smartphone cameras enable users to crowdsource images of the world around them. Cloud computing allows for cheap workload processing. Newer computer vision techniques allow 2-D images to be stitched together in a 3-D representation. Deep learning architectures improve the classification and segmentation of objects in an image.

Peter Kontschieder is the head of research at Mapillary, and he joins the show to talk about the technologies and research that has enabled Mapillary to build a futuristic business–an API for accessing information about the physical world.

Software Engineering Daily is looking for sponsors. If you are interested in reaching over 50,000 developers, you can go to softwareengineeringdaily.com/sponsor to find out more, and you can send us a message. We’d love to hear from you. And if you are an engineer working at a company that is marketing to developers, or hiring developers, if you tell your marketing department or your recruiting department about softwareengineeringdaily.com/sponsor, that is one way to help us out.

The post Computer Vision with Peter Kontschieder appeared first on Software Engineering Daily.

153 episodes available. A new episode about every 6 days averaging 57 mins duration .