Camio® Interest-based Video Compression™ is a way to vary the video encoding, resolution, storage, and analyses based on the interest-level of what's actually happening in the video event. For example, a person approaching the rear entrance of a business at 2am is more important to capture in HD, store in the cloud, and analyze fully than a leaf blowing in the wind.
Traditional video encoding works at the pixel level to reduce video to its most compact representation. But Camio's Machine Learning ranks the interest of each event in real-time to make the optimal choices for the most cost-effective video surveillance.
For example, the camera below has been trained to know that people approaching on the sidewalk are important and that all the passing cars on the street are not. Watch as the video switches from low resolution time lapse to full resolution as the man approaches.
Machine Learning automatically switches from timelapse → HD video when something interesting happens for the best bandwidth efficiency.
Another example is this transition from timelapse to HD as soon as people enter the scene: