As much as I love Google Earth, I use Google Maps at least as often due to the amazing GPS navigation that it provides in my car. Part of that is due to their work in Street View over the years and the data (and imagery) that it collects. Part of what they’ve developed is a system for automatically reading house numbers, and it’s an amazing system.
MIT Technology Review has a great overview of how the system works, including some great tid bits such as this:
…they assume that the building number has already been spotted and the image cropped so that the number is at least one-third the width of the resulting frame. They also assume that the number is no more than five digits long, a reasonable assumption in most parts of the world.
But the team does not divide the number into single digits, as many other groups have done. Their approach is to localize the entire number within the cropped image and to identify it in one go—all with a single neural network.
It’s quite a fascinating read and I encourage you to check it out. Google has pushed harder and taken Street View further than anyone imagined they would, and it’s exciting to think about where it might be in a few more years.