Robotic hands that mimic the capabilities of the human hand have cost $10,000 or more, and computer vision systems have worked only in highly structured environments on a very limited set of objects.
Therefore, our contribution should lead to more robust computer vision systems capable of working with real applications.
As Winter explained, the team decided to use a more traditional computer vision approach instead of using deep learning to classify these objects.
Those are, however, evolving toward more intelligence with computer vision and more, and crawling toward a proper "robot" status.
Bry admits top drone makers like DJI and Parrot could get more serious about computer vision auto-pilot.
Now, lvl5 thinks it has an even better, cheaper and more plentiful solution: computer vision software that extracts visual landmarks like stop signs and landlines, then aggregates the data into a kind of 3D map of the world that enables cars to triangulate their locations down to within an inch.
The $8 million from Atomico, Sequoia, LDV Capital, and PlayFair will be used to expand operations outside of its home base of Malmo, including a new office in San Francisco, and also to hire more staff, specifically computer vision engineers to supercharge the kind of data that can been gleaned from Mapillary's image library.
The team did the same with a live mouse, demonstrating how, with the correct polymer material and air pressure, the hand was capable of securely gripping an object without harming it — an act that would have required a far more complex array of computer vision and sensors in a more rigid gripper.
Adding more talent, and building out more connections into the computer vision ecosystem, could be coming at the right time for Magic Leap.
