AI Paper Atlas

Research map / 3D vision / Gaussian splatting

GH-NAF: Grid-Adaptive Hash-Level-Attended Neural Attenuation Fields for Discrepancy-Aware CBCT

Seong Je Oh, Ju Hwan Lee, Chae Yeon Lim, Donghwan Lee, Myung Jin Chung, Kyungsu Kim

CVPR 2026 · 3D vision › Gaussian splatting · approach: Sparse-view Gaussian splatting for novel views

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
use uncertainty-weighted projection loss with discrepancy rendering instead of plain MSE (ablation)GH-NAF · Synthetic CTPSNR: 34.06 -> 39.82Improved
add hash-level attention alignment loss Lattn (ablation)GH-NAF · Synthetic CTPSNR: 39.59 -> 39.82Improved
add occlusion regularization Locc (ablation)GH-NAF · Synthetic CTPSNR: 39.75 -> 39.82Improved
add density regularization Lden together with Locc (ablation)GH-NAF · Synthetic CTPSNR: 39.28 -> 39.82Improved
use GH-NAF instead of NeRF baselines under strong projection discrepancyNAF/SAX · Chest phantomMANIQA: SAX 0.353 -> 0.392 (150 views)Improved
use GH-NAF instead of Gaussian methods on discrepancy-free dataR2-Gaussian · Synthetic CT (15 views)PSNR: 34.34 -> 33.56Hurt
use GH-NAF instead of Gaussian methods on discrepancy-free dataR2-Gaussian · Synthetic CT (15 views)SSIM: 0.895 -> 0.934Improved
use GH-NAF instead of R2-Gaussian on low-discrepancy FIPSR2-Gaussian · FIPS (Walnut/Seashell/Pine)PSNR: Walnut 25v 45.24 -> 46.54; Seashell 25v 41.90 -> 39.91Mixed

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy