Going from scenes to the whole film in the future.

Posted: January 14, 2016 at 3:35 pm

I was thinking about having to use a small portion of the film during the residency and thinking about the feasibility of (ever) using the whole film. I thought of treating each scene as a separate unit and generating percepts and predictors on each separately.

There would then be a master predictor trained on the sequence of the scenes themselves. The master predictor would have two inputs and two outputs, one input for the ID of the scene, and one input for the length of the scene. That way the network would determine both the order of scenes and also the length of those scenes; Unfortunately, that would mean the master predictor would have to exploit each scene’s learning so each individual scene would still be the same every time.