Self-driving car technology faces a crucial test in the U.S.
Google has upgraded its Street View cameras to see the world better. It will feed such photos to its machine learning technology as a way to index the world and provide better, more nuanced location info. ( Google )
For the first in about eight years, Google has finally upgraded its street view cameras — yes, those globe-shaped shooters mounted on top of its self-driving cars that take street pictures as it travels the whole world.
The new camera setup will be smaller, now down to seven cameras instead of 15, all of which are 20-megapixel sensors. These, according to a report by Wired, will take clearer, higher-resolution street view photographs, in addition to having more vivid colors. It also includes two cameras that take still HD photos, along with two “cans” on both the front and back for laser radar.
All that’s mentioned above, however, are just means to an end. These higher-res cameras only serve to help Google’s machine learning technology. Photos of streets, highways, roads, and all sorts of pathways will help the company build a sort of mapped index of the world — in ways that will offer more information than a map will ever be able to give.
In short, thanks to the upgraded cameras, machine learning will have better pictures to work with, and in turn, it will be able to do its algorithm magic more efficiently. It will piece together information present in those photographs — identifying business names, street signs, and maybe even opening and closing hours of stores.
Google Street View And Machine Learning
What’s machine learning for, anyway? Well, Google says location queries have gotten tougher these days. In the early heydays of Google Maps, one might have only searched for their own address and several famous landmarks such as the Eiffel Tower. Now, they’ve gotten a lot more specific. For instance, a person wants to know “a Thai place open now that does delivery to my address,” which is far less specific than looking for the Stonehenge.
In the future, thanks to machine learning, Google VP and mapping head Jen Fitzpatrick says Google Maps might be able to answer more complex questions about certain locations such as, “What’s the name of the pink store next to the church on the corner?” Fitzpatrick says the only possible way to answer that kind of question is by having richer and deeper information.
Google’s machine learning tech has improved a lot since Street View cameras started taking photos in 2007. It has learned to figure out shorthands, as in “AV” stands for “avenida,” Portuguese for avenue.
Google Street View Rigs
Google’s street view team not only includes cars, but snow mobiles, tirkes, and a trolley, as well. There are even special rigs that can be snapped onto a backpack, mainly used to index places inaccessible to cars, such as narrow pathways.
© 2017 Tech Times, All rights reserved. Do not reproduce without permission.