DARPA's video search push gets another $11m
Not content with giving Google just one business model
Pentagon R&D chiefs at DARPA have awarded $11m to discover a technical secret for which, one may be sure, Google executives would pay a substantially larger sum - that of true video search.
Google would like to make video searching work as well as text searching does, the better to serve ads alongside the ensuing results and as a way of making its huge pile of video gumble at YouTube more useful and hence more valuable. But the problems of sifting the odd decent/useful vid out from among endless footage of comical cats, doleful teenagers in their bedrooms etc are as nothing compared to those of the US Defense Department.
The US military (and intelligence community) have been generating video archives over the past decade which make YouTube look like a home video collection. Mere HD movies and TV are small and tightly compressed compared to the high resolution, full-motion imagery which pours in such mighty streams from every Predator or Reaper roboplane: and dozens of these are airborne above southwest Asia every minute of every day.
Often, all these mountains of super-quality video (often featuring mountains, appropriately enough) are seen just once, briefly, by the pilot and sensor operator handling a given drone via satcomms from Nevada. In future, as the machines become more automated, large amounts of perhaps useful video will never be seen by human eyes at all: there is simply too much to assign intelligence analysts to watch it all in the hope of catching something significant.
Currently, video analysis for Predator and other aerial video surveillance platforms is very labor intensive, and limited to metadata queries, manual annotations, and “fastforward” examination of clips.
Or in other words the US military has bitten off more video than it can chew here. DARPA's solution, of course, is basically to invent a tremendous, powerful pair of robotic chewing jaws with steely automatic teeth. Rather than VID-CHOMP*, for some reason, DARPA has chosen to dub the project Video and Image Retrieval and Analysis Tool (VIRAT):
The software tools developed under VIRAT will radically improve the analysis of huge volumes of video data by: 1) alerting operators when specific events or activities occur at specific locations or over a range of locations and; 2) enabling fast, content-based searches of existing video archives.
VIRAT has been underway since 2008, and there have been previous awards: but on August 27 the Pentagon notified that New York firm Kitware Inc had received a $10,962,069 add-on to a previous cost-plus deal - indicating that DARPA feel worthwhile progress is being made.
Just as DARPA gave birth to the internet and thus to Google's original business model, so it seems that the military boffins may in time build the search colossus a new one.
Watch this space, etc. ®
*Video Intelligence Data - Computing for Human Observation Minimisation Programme
Is anyone else terrified by this?
If this technology exists and at some point in the future Labour gets elected, the UK is humped. Soon there after I am sure Australia will fall. After that, Canada can't be far behind. THIS MUST BE STOPPED, FOR THE GOOD OF US ALL.
Share Slow Motion Sports VIDEO - Instantaneously
I've been doing digital video for almost 8 years now and can randomly search and display any portion of MPEG videos. Having complete control of your video makes it usefull. Nobody has better control of video than this.
Here's what the software does with GOLF video - It works equally well for Football and other action sports.
The initial plan was to shoot video of each golfer from behind and across; then play these short (4 to 6 secs) videos in SLO-MO and NORMAL speed before the awards luncheon started.
This was ACCOMPLISHED and the 180+ videos (90+ participants and 2 CamCorders) were playing on a large screen within 10 minutes of leaving the course. The only PROBLEMS were: that a few golf swings took far more time than others (10 to 15 secs) and when played back in slow motion; the audience would lose interest. As well; in attempting to capture just the "golf swing" we missed some completely.
The SOLUTION was a program fix that allowed for only the FINAL 3 secs of the videos (the swing) to be played in slow motion. With this change; we can start recording when the golfer addressed the ball and shut it off on completion of the swing. No more missed swings or bored viewers.
Now they see each video clip followed by the final 3 seconds in Slo-Mo, as a screen saver.
One FINAL problem: After the meal there was a silent auction. Golfers wanted to see their swing (2 clips out of nearly 200). Again the solution came from the record on/off button. To achieve this and not lose any time by having to manually edit the text in the catalog file. An extremely short video clip (less than 2 Secs.) is created as each group comes to the Tee; of a white-board containing the golfer names and flight#. The software identifies these mini files and creates break points in the catalog file. The golfer tells us which group they were in, from the tournament sheet and we play that foursome's videos based on the Break Points (eg the 12th group we videoed). Upon viewing those videos we determine the golfers sequence # and select it (ie "Grp#12" "Seq#3"). So their videos from camcorder #1 then camcorder #2 play over and over again. The combination of these enhancements enables synchronizing video from 2 or more camcorders. What a BONUS.
The 180+ videos fit on 2 GigaByte USB sticks. These sticks are quickly loaded with the videos & software before the golfers leave. They just LOVED the immediacy.
SEE the above link; for video clips from those 2 tournaments. 20 of the 360+ videos are being switched out daily.
This method is extremely easy; because NO EDITING is required. Off-load the MPEG video from the 2 camcorders (5 to 10 mins). Catalog the video to play normal speed and Slo-mo (mere seconds). Then start the replay. VERY timely. All golfers from the rank amateur to the seasoned pro; want to see golf swings in Slo-Mo. Having hundreds of dual-speed golf swings playing randomly by foursome on the clubhouse screen is far more interesting to the patrons.
This changes the way we view video forever.