radiance-fields

Source code for WACV 2026 paper "WarpRF: Multi-View Consistency for Training-Free Uncertainty Quantification and Applications in Radiance Fields"

3D Gaussian Splatting training on Apple Silicon. Metal kernels that numerically reproduce gsplat — no Python, no CUDA.

Turn a video into an explorable 3D Gaussian Splat — fully local on Apple Silicon

AURA — Adaptive Unified Radiance Asset: the post-3DGS asset layer. Typed carriers (Gaussian/Beta), calibrated + certified per-carrier confidence exported in KHR_gaussian_splatting & OpenUSD 26.03, unified ray-query, relight preview.

Real-time 3DGS studio for lenticular light-field displays.

RadianceKit — native macOS app for 3D Gaussian Splatting (local, Apple Silicon/Metal). Open docs: export-format reference, capture guide & FAQ.

Capture reality with your phone, explore it as a real-time photorealistic 3D scene — browser-based Gaussian Splatting viewer. WebGL performance engineering, fully client-side.

DisplayXR demo — real-time 3D Gaussian Splatting viewer for glasses-free 3D displays (OpenXR + Vulkan, Windows & macOS).

Differentiable Gaussian rasterization with depth & alpha output — fork of graphdeco-inria/diff-gaussian-rasterization, used for few-shot 3DGS.

Few-shot 3D Gaussian Splatting for sparse-view (DTU) reconstruction, using an RGBD patch-diffusion prior (DiffusioNeRF) and PatchmatchNet MVS initialization.

Drop-in, idempotent speed patches for PGSR surface-reconstruction training — removes per-iteration PCIe transfers, redundant ones-kernel convolutions, and GPU-CPU logging syncs (plus a real EMA-logging bug) with CPU-proven numerical parity: same geometry, lower GPU bill.

Denoising-diffusion RGBD patch prior for regularizing radiance fields — fork of nianticlabs/diffusionerf, used as the diffusion regularizer for few-shot 3DGS.

Photos to 3D on a free Colab GPU: CUDA-enabled COLMAP matching, a desktop app for camera positions, and Gaussian Splatting training. 320 photos matched in 28 minutes instead of 11 hours.