The web is supposedly full of useless content, but sometimes you just need to connect the dots to create something amazing. That's exactly what Google and the University of Washington have done by building a platform that analyzes, processes, and coordinates millions of public images to produce stunning time-lapse videos.
Time Lapse Mining
Coincidences always raise suspicions in the world of computing, no matter how small they are. We recently told you about Hyperlapse, a new Microsoft application for creating hyperlapse videos from your own material, and now it's Google's turn. The Mountain View giant holds a privileged position on the web, and when it comes to finding content, it knows how, where, and when to do it. In early August, the 2015 edition of the SIGGRAPH conference will take place, and that's the venue chosen by Google to present a new system developed jointly with researchers from the University of Washington.
The official term is time lapse mining, and this system collects publicly accessible images from various services (Picasa, Flickr, etc.) to synthesize small time-lapse videos of specific locations.
How It Works
In the demonstration, 86 million photos were used, processed to compensate for differences in color and lighting. They also worked on reducing the flicker in transitions (caused by variables in light sources), and finally all frames are stabilized, avoiding unwanted vibrations in the final video. Buildings under construction, glaciers shrinking, waterfalls with altered water courses, and even archaeological excavations are part of this impressive demonstration.
Release Plans
Fortunately, the code behind these videos won't remain in private hands. Both Google and the university plan to release it soon, though I'd guess it will be after the SIGGRAPH 2015 presentation. The number seems outrageous at first glance, but 86 million images are actually a tiny fraction of what's floating around on the web. How many more videos could be designed with the remaining content...?
