Multi-view Supervision for Single-view Reconstruction via Differentiable\n Ray Consistency

We study the notion of consistency between a 3D shape and a 2D observation\nand propose a differentiable formulation which allows computing gradients of\nthe 3D shape given an observation from an arbitrary view. We do so by\nreformulating view consistency using a differentiable ray consistency (DRC)\nterm. We show that this formulation can be incorporated in a learning framework\nto leverage different types of multi-view observations e.g. foreground masks,\ndepth, color images, semantics etc. as supervision for learning single-view 3D\nprediction. We present empirical analysis of our technique in a controlled\nsetting. We also show that this approach allows us to improve over existing\ntechniques for single-view reconstruction of objects from the PASCAL VOC\ndataset.\n