SDF-SRN: Learning Signed Distance 3D Object Reconstruction from Static\n Images

Abstract

Dense 3D object reconstruction from a single image has recently witnessed\nremarkable advances, but supervising neural networks with ground-truth 3D\nshapes is impractical due to the laborious process of creating paired\nimage-shape datasets. Recent efforts have turned to learning 3D reconstruction\nwithout 3D supervision from RGB images with annotated 2D silhouettes,\ndramatically reducing the cost and effort of annotation. These techniques,\nhowever, remain impractical as they still require multi-view annotations of the\nsame object instance during training. As a result, most experimental efforts to\ndate have been limited to synthetic datasets. In this paper, we address this\nissue and propose SDF-SRN, an approach that requires only a single view of\nobjects at training time, offering greater utility for real-world scenarios.\nSDF-SRN learns implicit 3D shape representations to handle arbitrary shape\ntopologies that may exist in the datasets. To this end, we derive a novel\ndifferentiable rendering formulation for learning signed distance functions\n(SDF) from 2D silhouettes. Our method outperforms the state of the art under\nchallenging single-view supervision settings on both synthetic and real-world\ndatasets.\n