3D Gaussian Splatting
A real-time 3D rendering technique that gained wide adoption after 2023, known for its high rendering speed and visual quality. The approach represents scenes using a collection of 3D Gaussian primitives, substantially lowering the technical barrier for real-time interactive scene rendering compared to earlier neural rendering methods.
Core Facts
NeRF (Neural Radiance Fields) and its successor Gaussian Splatting can reconstruct three-dimensional scenes that can be freely navigated from any viewpoint from 2D images.
3D Gaussian Splatting (3DGS) was proposed in 2023 by the French INRIA team, achieving rendering speeds dozens of times faster than NeRF while maintaining comparable image quality.
3D Gaussian Splatting elevated neural rendering speed to real-time levels in 2023.
Timeline (last 90 days)
VisibilitySplat proposes a method that retains the efficient training and visualization pipeline of rasterized 3DGS while additionally attaching a sparse visibility "oracle" to handle occlusion queries.
3DGS uses rasterized splatting, in which mutual occlusion between different Gaussian primitives is not explicitly modeled.
3DGS depicts static geometry and appearance at a single moment and inherently contains no temporal dimension; making a scene move requires introducing time-varying deformation or an additional 4D representation.
SplatQ is a combinatorial optimization pruning algorithm for 3D Gaussian Splatting (3DGS) that coordinates regional budget allocation and joint Gaussian selection under a fixed budget.
Compared to NeRF, which requires dense ray-marching sampling, 3DGS uses a rasterization pipeline that is better suited for GPU parallelism and can achieve real-time frame rates.
Global Top-K pruning has two structural defects: lack of regional budget control and neglect of joint deletion effects.
Most existing mainstream 3DGS pruning methods assign independent importance scores to each Gaussian and perform global Top-K selection.
3DGS is one to two orders of magnitude faster than NeRF in volumetric rendering with comparable quality.
Traditional 3DGS relies on a complete SfM point cloud for initialization and cannot be directly applied to incremental online scenes.
The mainstream approaches to AI avatar technology fall into two categories: diffusion-model-based image animation and 3D reconstruction based on NeRF or 3D Gaussian Splatting.
14 more timeline events
All Facts (20)
NeRF (Neural Radiance Fields) and its successor Gaussian Splatting can reconstruct three-dimensional scenes that can be freely navigated from any viewpoint from 2D images.
80%Verified3D Gaussian Splatting (3DGS) was proposed in 2023 by the French INRIA team, achieving rendering speeds dozens of times faster than NeRF while maintaining comparable image quality.
80%Verified3D Gaussian Splatting elevated neural rendering speed to real-time levels in 2023.
75%Verified3DGS explicitly represents scenes using millions of three-dimensional Gaussian ellipsoids with attributes including position, covariance matrix, color (spherical harmonic coefficients), and opacity.
65%Verified3D Gaussian Splatting uses explicit Gaussian spheres as basic units, achieving a substantial increase in rendering speed, and has rapidly gained adoption in real-time applications in recent years.
65%UnverifiedCompared to NeRF, 3DGS training is several to more than ten times faster, real-time rendering frame rates can reach tens to hundreds of FPS, and scenes can be directly exported as point cloud assets.
60%UnverifiedThe neural rendering technical path, represented by NeRF (Neural Radiance Fields) and 3D Gaussian Splatting, has the potential to bypass the industrial pipeline of traditional polygon modeling.
60%UnverifiedVisibilitySplat proposes a method that retains the efficient training and visualization pipeline of rasterized 3DGS while additionally attaching a sparse visibility "oracle" to handle occlusion queries.
50%Unverified3DGS uses rasterized splatting, in which mutual occlusion between different Gaussian primitives is not explicitly modeled.
50%Unverified3DGS depicts static geometry and appearance at a single moment and inherently contains no temporal dimension; making a scene move requires introducing time-varying deformation or an additional 4D representation.
50%UnverifiedGlobal Top-K pruning has two structural defects: lack of regional budget control and neglect of joint deletion effects.
50%UnverifiedCompared to NeRF, which requires dense ray-marching sampling, 3DGS uses a rasterization pipeline that is better suited for GPU parallelism and can achieve real-time frame rates.
50%UnverifiedMost existing mainstream 3DGS pruning methods assign independent importance scores to each Gaussian and perform global Top-K selection.
50%UnverifiedSplatQ is a combinatorial optimization pruning algorithm for 3D Gaussian Splatting (3DGS) that coordinates regional budget allocation and joint Gaussian selection under a fixed budget.
50%Unverified3DGS is one to two orders of magnitude faster than NeRF in volumetric rendering with comparable quality.
50%UnverifiedTraditional 3DGS relies on a complete SfM point cloud for initialization and cannot be directly applied to incremental online scenes.
50%UnverifiedThe mainstream approaches to AI avatar technology fall into two categories: diffusion-model-based image animation and 3D reconstruction based on NeRF or 3D Gaussian Splatting.
50%UnverifiedA typical 3DGS scene often requires millions to tens of millions of Gaussians, with per-scene storage reaching hundreds of MB.
50%Unverified3D Gaussian Splatting (3DGS) is a real-time neural scene rendering method proposed in 2023 that represents scene geometry and appearance using a large number of 3D Gaussian ellipsoids.
50%UnverifiedExisting 3DGS technical approaches universally treat each element in the scene as a unique entity, storing a large number of redundant parameters and providing very limited manipulation capability for downstream editing tasks.
50%