LiveLight Uses Machine Learning to Cut the Boring Parts of Film Reels

By: Alyson Wyers - Aug 1, 2014
References: cmu.edu & springwise
This self-editing video uses machine learning algorithms to cut out the boring parts of security footage and consumer videos. Developed by Carnegie Mellon University in Pennsylvania, LiveLight can automatically edit film reels to create a montage of interesting moments.

The self-editing video technology contains a 'learned dictionary' that can be taught to recognize repetitive moments and limited activity and cut it out. This algorithm allows LiveLight to scan footage and pick out events, which it compares and contrasts to keep the more active or new sequences. Springwise uses the example of a long static shot of cars driving by. It would be reduced to a shorter clip because it's repetitive, versus a car accident which would be recognized as a new event.