Smart tagging at REAL2016

Posted By Scott Singer on March 21, 2016 06:53 am | Permalink

At Autodesk's REAL2016 conference, there was a bit of technology that might become of some use for film and video editorial ingest. The conference, which brings together reality capture and computing technology to service the architecture, industrial design and civil engineering communities, often overlaps with other fields. It turns out that the construction industry also needs to sift through hours of source footage - in their case from construction sites looking for views of progress, installations and the like. 

A company called Smartvid (smartvid.io) has come up with a way, using deep learning and neural networks, to auto-classify construction site video; it does this using keyword tags and other metadata. Founder and CEO Josh Kanner, says that while currently the system works on the audio track of the video where a user interactively "tags" the session with words like "ductwork," in the future they are working toward incorporating computer vision technology to classify and tag video directly from the image streams. And while that's a future application, he and I talked about the ability currently, to record an alternate audio track over the source footage to tag the footage for editorial purposes, essentially live tagging the who, what, when, where and why, of shots as they're viewed in editorial for automatic classification.

Definitely technology to keep an eye on. #smartvid.io #deeplearning #neuralnetwork