Reproduce “D2NT: A High-Performing Depth-to-Normal Translator” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Yi Feng, Bohuan Xue, Ming Liu, Qijun Chen, Rui Fan
A novel Depth-to-Normal Translator (D2NT) enables real-time surface normal estimation with enhanced accuracy, particularly near discontinuities, by integrating a discontinuity-aware gradient filter and an MRF-based refinement module. This system achieves a runtime as low as 1.82 ms and average angular errors down to 0.89° on challenging datasets.
Key insights
- A direct mathematical formulation can efficiently translate depth to surface normals, eliminating the overhead of 3D point cloud reconstruction.
- Adaptive gradient filters, capable of distinguishing continuous surfaces from outliers, are critical for accurate normal estimation around discontinuities.
- A modular, MRF-based refinement approach can serve as a universal backend optimizer, substantially improving the accuracy of existing depth-to-normal estimators.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.