Just about every device has a camera in it, so we’re shooting more and more video than we have ever before. If only all of it were worth watching.
The latest application of machine learning was developed by Eric P. Xing, professor of machine learning from Carnegie Mellon University, and Bin Zhao, a Ph.D. student in the Machine Learning Department. It’s called LiveLight, and it can help automate the reduction of videos to just their good parts.
To get a quick idea what it’s all about, watch the demo above.
LiveLight takes a long piece of source footage and “evaluates action in the video, looking for visual novelty and ignoring repetitive or eventless sequences, to create a summary that enables a viewer to get the gist of what happened.” Put another way, it watches your movie and edits out the boring stuff. This all happens with just one pass through said video — LiveLight never works backwards.
You’re left with something more like a highlight reel than the too-long original video pictured on the left above. LiveLight is robust enough to run on a standard laptop and is powerful enough to process an hour of video in one or two hours.