Big Data Takes on Video — And It’ll Change The Medium for Good

Big data, like any trending topic with the ability to rapidly transform our lives, is evolving. Simply taking a quick look at the latest headlines will prove that:

See, big data is no longer an incomprehensible tech term. It’s turning into policy, business, and the maker and shaker of careers, markets and more. Further, big data is making its way from the web to the Internet of Things, where mobile and specifically video is proving to be more valuable than ever before. 

We aren’t talking meta-tagging of videos. This is bigger than that. Sure, you can search on YouTube for a video with a hamster eating a burrito, and what you’re looking for will be surfaced thanks to keywords based on title and those that were input when the video was uploaded. Then again, you’ll also get 2,840 other results — to be exact — with similar keywords.

“Big data is no longer an incomprehensible tech term. It’s turning into policy, business, and the maker and shaker of careers, markets and more.”

Object recognition software, though, much of which is already in testing, will eliminate the need for manual tags or keywords, and hence, eliminate thousands of irrelevant search query results. We’re talking about algorithms that scour the frames of each video and build a description of what is going on — all on their own.

This isn’t a novel implementation though — and most of us are already using such technology on a daily basis. 

On Facebook, for instance, face recognition technology is now standard. The company’s DeepFace software creates 3D images of faces in pictures, rotates them as if they were looking at the camera and then cross-checks to see if there are enough similarities to suggest a tag. And, with 97.25% accuracy, DeepFace nearly matches the human brain’s accuracy for facial recognition.

Soon, that same accuracy will be true on video as well. Researchers at the University of Texas at Austin are working on algorithms that determine specific objects in videos, creating short descriptions for videos without using keywords, tags or titles — and they are doing it by analyzing the each frame of a video. 

So what are some of the benefits of gathering data from video? Prism Skylabs, a cloud service, allows retail companies to utilize in store video monitoring to gather data on customer tendencies, including what products are being picked up the most and where the most traffic is seen in the store. Utilizing raw video to gather data is helping companies understand how their customers interact with their business, creating a better customer experience leading to increased loyalty and ROI. 

While in its infancy, video object recognition will soon be standard — and the big data usages that come from it will be much more valuable than current engagement and impressions metrics. Whether it is through Google Glass, in-store video monitoring systems or a professional camera court-side at a NBA playoff game, the data entry points for big data and video are infinite. And it will change the way we see the world through a camera’s lens.