A US-based startup has successfully built a new tool that will give anyone and everyone the power to scan the entire Earth through Artificial Intelligence (AI) “eyes” and instantly find satellite images of matching objects. This means, if you want to go on a world tour of all the major cricket stadiums on the planet, the tool will come to your help by churning out your itinerary for the same.
The tool named GeoVisual Search has been conceived and developed by a New Mexico-based startup Descartes Labs, which provides Artificial Intelligence-driven analysis of satellite images to academics, industry and governments. The startup unveiled its newest tool, GeoVisual Search to the world by releasing a public demo of how it works. The tool is basically a new type of search engine that successfully combines satellite images of planet Earth with machine learning on a large scale.
The GeoVisual Search tool allows its users to pick any object of their choice anywhere on Earth (the only condition being it should be visible from space), and then use the tool to churn out a list of similar-looking objects along with their locations on the planet.
Tech enthusiasts might be quick to point that the tool sounds awfully close to Google Earth, but fortunately, that is not the case at all. While Google Earth just allows its users to search geotagged locations around the world, the GeoVisual Search tool, on the other hand, actually compares all the pixels to make up huge photographs of the world in order to find us the results of all the matching objects, something which wasn’t possible for the masses before on such a global scale.
According to a statement given by Mark Johnson, Descartes Lab CEO and co-founder, their goal with GeoVisual Search is to make more and more people aware about the huge potential of machine learning. They aim to use this data to model complex planetary systems, and according to Johnson, the tool is just the first step in that direction. He’s hopeful that the tool will stir a discussion among businesses on how they can make use of new kinds of data to improve their work and efficiency.
Descartes Labs still has to work out some quirks from the tool. Also, since GeoVisual Search operates on top of an intelligent machine-learning platform, it can be trained and will improve over time.
Currently, the demo is dependent on three different imagery sources that is inclusive of more than 4 petabytes of data altogether. One can make use of the National Agriculture Imagery Program (NAIP) data to search for the lower 48 United States because it boosts of having the highest resolution of one meter per pixel, which makes it capable of spotting solar farms, turbines, among many other objects.
The tool has four-meter imagery available for China that makes it possible to recognise slightly larger things like cricket stadiums etc. As far as the rest of the world is concerned, the startup is currently making use of 15-meter resolution images from Landsat 8 which are a little coarse but they still allow for the tool to identify larger-scale objects.
Tempted to try out the tool, you can do it right here.