RGBD-to-3D Object Mesh Refinement via Depth Matching and Symmetry Propagation

Key points

• Single-view 3D reconstructors often produce plausible meshes that disagree with the input view, especially near depth discontinuities and self-occlusions.1
• Given a depth map, we correct the visible surface by bipartite matching to back-projected depth points, mirror these corrections onto the occluded side across a detected symmetry plane, and propagate them with a smoothness solver.1
• On GSO and OmniObject3D with five backbones, it yields consistent gains, also with monocular pseudo-depth, benefits more from symmetry on symmetric objects, and compares favorably with prior refinement in accuracy and runtime.1
• It further improves an RGB-D-to-mesh reconstructor and transfers to real captures with noisy sensor depth.1

Sources (1)

• [1]RGBD-to-3D Object Mesh Refinement via Depth Matching and Symmetry Propagation
  arXiv (AI, ML, NLP, CV, robotics, multi-agent) · Oct 8, 03:44 AM

  We present a lightweight, plug-and-play RGBD-to-3D refinement that improves any RGB-to-3D reconstructor without retraining.

  Single-view 3D reconstructors often produce plausible meshes that disagree with the input view, especially near depth discontinuities and self-occlusions.

Extractive summary: sentences quoted from the sources.