Geometry Regularization for Vanilla NeRF via Self-Supervised Cross-Ray Consistency

Key points

• However, the geometry encoded by the density field of vanilla NeRF can be noisy and inaccurate, since different combinations of shape and radiance may explain the same input images.1
• To reduce this ambiguity, we augment vanilla NeRF with a geometry-optimization branch based on self-supervised cross-ray consistency.1
• Based on the visibility-weighted color consistency across these rays, we constrain the implicit geometry.1
• We further regularize the implicit geometry by penalizing opacity at candidates with inconsistent appearance predictions and encouraging local normal smoothness.1

Sources (1)

• [1]Geometry Regularization for Vanilla NeRF via Self-Supervised Cross-Ray Consistency
  arXiv (AI, ML, NLP, CV, robotics, multi-agent) · Oct 20, 07:40 AM

  Neural Radiance Fields (NeRF) achieve high-quality novel-view synthesis by representing a scene as a continuous volumetric radiance field.

  However, the geometry encoded by the density field of vanilla NeRF can be noisy and inaccurate, since different combinations of shape and radiance may explain the same input images.

Extractive summary: sentences quoted from the sources.