Almost two years ago I wrote a post imagining a future where intelligent asset management software would remove most of the labor involved in managing large image collections. Fortunately, we’re starting to see gains in image recognition software which promise to make solving these difficult tasks easier and less time consuming. I recently learned of some new technologies that are worth a mention, potentially turning the visual and writing world on their heads.
One of the primary uses of the PFTrack application is to track motion in video files to ensure composited material matches and moves with footage captured in-camera. After the movement within an image is tracked, and points of detail and contrast are analyzed, the software, (with a little manual help), can then infer distance and depth to place those points in three dimensions. These features provide two separate ways of extracting three dimensions from two dimensional video footage.
A second use of this technology, shown in a demo from FXGuideTV (see 18:00 to 21:43), PFTrack analyzes dozens of points from a moving video capture of a face, or a landscape and builds a 3d model based on this data.
The possibilities inherent in combining this type of deep image analysis with large, photographic data sets like Google Earth or Bing Maps are remarkable. Right now, the software requires significant manual intervention by a trained user, but it stands to reason that within several years, we could conceivably have a comprehensive, three dimensional, photo-realistic set of maps to repurpose in a number of different ways, from navigation to education.
Automation and intelligent software, can be used in other disciplines as well. Keith Gordon of Mullen advertising agency pointed me toward an NPR article where a robot sportswriter wrote a better synopsis of a collegiate baseball game than its human counterpart. The creators of the robotic-article-writing software, Narrative Science, hope to someday win a Pulitzer-prize for reporting. Given the massive amounts of data their software could sift through, from cancer rates and air quality statistics, to political donors and senatorial votes, it seems likely that at some point in the not-to-distant future, intelligent software could break a really big story that would have been impossible for a human to crack.