Gaussian Splat Engine

Modern C++ implementation of 3D Gaussian Splatting with CUDA acceleration.

Overview

An implementation of 3D Gaussian Splatting for educational and research purposes.Built with modern C++17, following the algorithm from the original paper with implementation details referenced from the official open-source code.

Features

• CUDA-accelerated Rasterization - Tile-based rendering for real-time performance
• Modern C++ Design - Clean architecture with C++17 features
• Training Pipeline - Full forward/backward implementation with optimization

Project Status

🚧 Work in Progress

Currently implementing:

• Project structure
• PLY file loader
• Visualization (1) (3D Color Points only)
• CPU rasterizer (for test)
• CUDA rasterizer (forward pass)
• Visualization (2) (Rendering)
• Training pipeline (backward pass)
• Optimization algorithms (Adam, SGD)

Architecture

Pure rendering and training engine

gaussian_splat_engine/
├── src/                    # C++ Core Engine
│   ├── database/           # Gaussian data structures
│   ├── rendering/          # Forward pass (CUDA)
│   ├── training/           # Backward pass (CUDA)
│   ├── optimization/       # Optimizers (Adam, SGD)
│   └── util/               # PLY loader, helpers
│

Design Philosophy:

• Core engine in C++/CUDA for performance
• Modular architecture for flexibility
• Minimal dependencies

Build Requirements

• CUDA 11.6+
• CMake 3.18+
• C++17 compiler (GCC 9+, Clang 10+)
• Eigen3

⚡ Built with performance in mind | 🎨 Designed for clarity | 🚀 Optimized with CUDA

---

License & Attribution

This project is for research and educational purposes only.

Original 3D Gaussian Splatting

This implementation follows the algorithm described in:

3D Gaussian Splatting for Real-Time Radiance Field Rendering
Bernhard Kerbl, Georgios Kopanas, Thomas Leimkühler, George Drettakis
ACM Transactions on Graphics, 2023
Project Page | Paper

Implementation details referenced from the official open-source code:
graphdeco-inria/diff-gaussian-rasterization

Citation

If you use this code in your research, please cite the original work:

@Article{kerbl3Dgaussians,
      author       = {Kerbl, Bernhard and Kopanas, Georgios and Leimk{\"u}hler, Thomas and Drettakis, George},
      title        = {3D Gaussian Splatting for Real-Time Radiance Field Rendering},
      journal      = {ACM Transactions on Graphics},
      number       = {4},
      volume       = {42},
      month        = {July},
      year         = {2023},
      url          = {https://repo-sam.inria.fr/fungraph/3d-gaussian-splatting/}
}

This Implementation

This repository provides a CPU-based educational implementation for learning purposes.

Author: Seungwon Choi ([email protected])
Purpose: Research benchmark for monocular Gaussian splatting SLAM