Federico Zertuche

We try to reconstruct a structured 3-d scene from a single still image (photograph). The method is heavily inspired in how a human sees. "Seeing is a collection of inferences about the world. Motion, color and depth are important individual judgments. To see, however, we must connect these inferences into a unified explanation of the image.  The algorithm we will study does just that: It infers monocular individual judgments and connects them using a probabilistic model.