Shape Recipes: Scene Representations that Refer to the Image

The Problem: The goal of low-level vision is to estimate an underlying scene (e.g., albedos or shapes), given an observed image. Real-world scenes can be very complex, conventionally requiring high dimensional representations that are hard to estimate and store. In this work, we propose a low-dimensional representation, called a scene recipe, that relies on the image itself to describe the complex scene configurations. Current work focuses on shape recipes that are a particular case of scene recipes in which the quantity of interest is the 3D-shape that corresponds to an image.