Eyeful Tower Comparisons show that HybridNeRF outperforms MERF, 3DGS, and VolSDF in HDR training and exposure adjustments. The implementation using iNGP acceleration primitives allows for rendering over 10x faster than VR-NeRF with slightly better quality. This innovative approach excels in modeling specular effects, surpassing faster rendering methods. The authors Turki, Agrawal, Bulò, Porzi, Kontschieder, Ramanan, Zollhöfer, and Richardt present their findings in “HybridNeRF: Efficient Neural Rendering via Adaptive Volumetric Surfaces,” published at CVPR 2024.
https://haithemturki.com/hybrid-nerf/