ICASSP 2011 Demo

Here is a work in progress demo showing some of the capabilities we hope to include in our final system, in this case we are generating our depth model and synthesising outputs completely from scratch with no human intervention using only the input images with no additional information.

From a set of 25 images arranged in a grid we take two corner images as our key images, generate disparity maps using the other images in the set and then synthesis new views for a virtual camera with the path shown in the video: