Exploiting Symmetry and/or Manhattan Properties for 3D Object Structure Estimation from Single and Multiple Images

Many man-made objects have intrinsic symmetries and Manhattan structure. By assuming an orthographic projection model, this paper addresses the estimation of 3D structures and camera projection using symmetry and/or Manhattan structure cues, for the two cases when the input is a single image or multiple images from the same category, e.g. multiple different cars. Specifically, analysis on single image case implies that Manhattan alone is sufficient to recover the camera projection, then the 3D structure can be reconstructed uniquely exploiting symmetry. But Manhattan structure can be hard to observe from single image due to occlusion. Hence, we extend to the multiple image case which can also exploit symmetry but does not require Manhattan axes. We propose a new rigid structure from motion method, exploiting symmetry, using multiple images from the same category as input. Our results on Pascal3D+ dataset show that our methods can significantly outperform baseline methods.