Fast Dynamic Radiance Fields with Time-Aware Neural Voxels

Text-To-4D Dynamic Scene Generation

“…Another category of approaches learn a time-varying deformation of 3D points into a static canonical scene (Pumarola et al, 2021;Park et al, 2021a;Tretschk et al, 2021;Park et al, 2021b). Some approaches accelerate NeRFs on dynamic scenes using explicit voxel grids (Fang et al, 2022) or tensor factorization (Cao & Johnson, 2023;Fridovich-Keil et al, 2023).…”

Robust Dynamic Radiance Fields

“…A recent line of work extends NeRF to handle dynamic scenes [18,22,37,50,51,54,68,73]. Although these space-time synthesis results are impressive, these techniques rely on precise camera pose input.…”

“…NSFF [37] DynamicNeRF [22] HyperNeRF [51] TiNeuVox [18] Ours Ground truth Table 3. Novel view synthesis results.…”

“…Recent work has relaxed these restrictions and can handle unstructured video input [1,2]. Recently, there have been many dynamic view synthesis methods for unstructured videos [22,23,37,50,51,54,68,73] and other new methods based on deformable fields [18]. However, these techniques require precise camera poses as input, typically estimated via SfM systems such as COLMAP [59] (bottom left of Table 1).…”

“…Plenoxels [21], Instant-ngp [43], TensoRF [10] NeRF -- [70], BARF [38], SC-NeRF [29] Dynamic scene NV [41], D-NeRF [54], NR-NeRF [68], NSFF [37], DynamicNeRF [22], Nerfies [50], HyperNeRF [51], TiNeuVox [18], T-NeRF [23] Ours art methods on several challenging datasets that typical SfM systems fail to estimate camera poses.…”