CF-Splat turns triangle-splatting appearance estimation into a closed-form linear least-squares solve
zhenjun_zhao · x · 2026-10-17
- Motivation: Existing splatting radiance-field methods rely on iterative nonlinear optimization over millions of coupled parameters — slow, GPU-memory heavy, and prone to local minima.
- Key insight: Given known mesh geometry and calibrated cameras, rendered pixel colors are linear in per-vertex spherical harmonic (SH) coefficients, so multi-view appearance estimation becomes a global linear least-squares (LLS) problem with no iterative optimization.
- Design: Adaptive SH regularization constrains under-supported high-order coefficients for a unique global minimizer; a triangle-structured matrix-free preconditioned conjugate-gradient solver scales to large scenes without building the sparse normal system.
- Results: On C3VD (endoscopic) and OmniObject3D, CF-Splat matches or beats iterative methods (3DGS, Triangle-Splat, Mesh-Splat) on PSNR/SSIM/LPIPS while skipping iterative optimization — suited to robot perception pipelines that already produce meshes.
More from Embodied
- Contact-UAN: MIT's residual world model fixes humanoid sim-to-real with 6.5 minutes of data — kevin_zakka · 2026-10-17
- ROS 2 LiDAR localization runs 390 m outdoors on handheld Livox MID-360 — rsasaki0109 · 2026-10-17
- Embodied Astra tops RoboMME at 92.1% — beating humans by just keeping full visual history in context — hankyang94 · 2026-10-17
- ROS weekly: ROSCon China kicks off in Shanghai, ROS By-The-Bay set for Oct 22 at Intrinsic — kscottz · 2026-10-17
- Simulated robotics eval: Claude Opus 5.5 preferred over Opus 5 in 14 of 16 comparisons — qinzytech · 2026-10-17
- FoundationVIO fuses IMU with foundation-model predictions to fix metric scale in feed-forward SLAM — zhenjun_zhao · 2026-10-17