Gaussian splatting and Similar

Gaussian Splatting

• Instantsplat: Unbounded Sparse-view Pose-free Gaussian Splatting in 40 Seconds
• The Rise Of 3D Gaussian Splatting: What Is It And How Is It Changing The Immersive Media Industry? — Magnopus
• 4D Gaussian splatting and Similar with time domain
• Gaussian splatting and Similar gsgen
• Room scale Gaussian splatting and Similar technique for single lens (#SLAM) Scene Capture and Reconstruction Gaussian-SLAM: Photo-realistic Dense SLAM with Gaussian Splatting (vladimiryugay.github.io)
• Mip-Splatting (niujinshuchong.github.io) reduced artifacts in Gaussian splatting and Similar
• Gaussian-SLAM: Photo-realistic Dense SLAM with Gaussian Splatting (vladimiryugay.github.io)
• GaussianDiffusion: 3D Gaussian Splatting for Denoising Diffusion Probabilistic Models with Structured Noise
  • logseq://graph/researchpapers?block-id=6579a51f-5e6d-4570-903f-9458f84e845f
• Gaussian SLAM rooms scale scanning
  • logseq://graph/researchpapers?block-id=6579a880-ce7f-4a79-b3d3-9135ff4348b3
  • Gaussian Splatting SLAM (rmurai.co.uk) is near real-time
• Paper page TRIPS: Trilinear Point Splatting for Real-Time Radiance Field Rendering (huggingface.co)
• Deblurring 3D Gaussian Splatting (benhenryl.github.io)
• huggingface/gsplat.js: JavaScript Gaussian Splatting library. (github.com)
• Gaussian splatting and Similar in Houdini
• Triplane Meets Gaussian Splatting: Fast and Generalizable Single-View 3D Reconstruction with Transformers (zouzx.github.io) understandable 3D and 4D from Gaussian splatting and Similar
• dynamic3dgaussians.github.io using a multi Motion Capture dome and Gaussian splatting and Similar for 6DOF Human tracking and SLAM capture
• LangSplat: 3D Language Gaussian Splatting

NeRFs

• MobileNeRF: This approach adapts NeRFs for mobile devices by exploiting the polygon rasterization pipeline for efficient neural field rendering. It achieves very fast rendering times (0.016-0.017s) but requires long training times.
• MobileR2L: This method uses a full CNN-based neural light field model with a super-resolution model in its second stage. It achieves real-time inference on mobile devices while maintaining high image quality, rendering a 1008x756 image of real 3D scenes in 18.04ms on an iPhone 13.
• Instant NGP (Neural Graphics Primitives): Developed by NVIDIA, this technique significantly speeds up the training and rendering of NeRFs, allowing for near-instantaneous scene reconstruction.
• Plenoxels (Plenoptic Voxels): This method replaces neural networks with a sparse 3D grid of spherical harmonics, enabling faster training and competitive quality compared to NeRFs.
• NGLOD (Neural Geometric Level of Detail): This approach combines neural implicit representations with explicit geometric representations, allowing for multi-resolution rendering and faster training.
• NeRF-MAE (Masked AutoEncoders for NeRFs): This technique applies the concept of masked autoencoders to NeRFs for self-supervised 3D representation learning, potentially improving generalization and efficiency.

NeRFs vs Hardware Acceleration

• old page, needs Update Cycle
• Neural Rendering and Its Hardware Acceleration: A Review (arxiv.org)

Paper	Neural Network Type	Residual Layer	Concatenation Layer	Suitability for Low-end Mobile Hardware
GIRAFFE	MLP, CNN	Required	Required	7
Render Net	MLP, CNN	Not Required	Required	6
Neural Voxel Renderer	MLP, CNN	Not Required	Required	5
Neural Volumes	MLP, CNN	Not Required	Required	5
NeRF	MLP	Not Required	Required	8
NeRF in the Wild	MLP	Not Required	Required	7
KiloNeRF	MLP	Not Required	Required	8
FastNeRF	MLP	Not Required	Required	9
Plenoctrees	MLP	Not Required	Required	8
Instant Neural Graphics Primitives	MLP	Not Required	Required	9
Scene Representation Networks	MLP	Not Required	Required	7
Extracting Motion and Appearance	MLP, CNN, Transformer	Required	Required	6
Instant 3D	MLP	Not Required	Required	8
Neural Point Cloud Rendering	CNN, U-Net	Not Required	Required	6
Deep Shading	CNN	Not Required	Required	6
Neural Reflectance Fields	CNN	Required	Not Required	7
Deep Illumination	GAN, U-Net	Not Required	Required	5
Common Objects in 3D	MLP, Transformer	Required	Required	7
GeoNeRF	Transformer	Required	Required	7
Gen-NeRF	Transformer	Required	Required	7

• playcanvas/supersplat: 3D Gaussian Splat Editor Gaussian splatting and Similar
• Long Volumetric Video Gaussian splatting and Similar
• AniGS Humans, Avatars , Character Gaussian splatting and Similar