easy-3dgs 0.1.0

A Python library that includes an end-to-end 3D Gaussian splatting pipeline without requiring the compilation of hundreds of dependencies.

easy-3dgs

A Python library for 3D Gaussian Splatting, designed to be easy to use and extend.

Features

• Modular Pipeline: The reconstruction process is broken down into a series of modular steps, including feature retrieval, pair generation, feature extraction, feature matching, reconstruction, and image undistortion.
• Extensible: Each step in the pipeline is designed to be easily replaceable with custom implementations.
• HLOC-Based: Leverages the power of Hierarchical Localization (HLOC) for robust and accurate 3D reconstruction.

Getting Started

Prerequisites

• Python 3.10 or later
• Git

Installation

Install from PyPI

You can install the library directly from PyPI using pip:

pip install git+https://github.com/Jourdelune/easy-3dgs.git
pip install git+https://github.com/cvg/Hierarchical-Localization.git

Install from Source

1. Clone the repository:

   git clone --recursive https://github.com/Jourdelune/easy-3dgs.git
   cd easy-3dgs
   

2. Install dependencies:

   pip install -e .
   pip install git+https://github.com/cvg/Hierarchical-Localization.git
   

Usage

The main entry point for the library is the ReconstructionPipeline class. Here's a simple example of how to use it:

from pathlib import Path
from easy_3dgs.pipeline import ReconstructionPipeline
from hloc import extract_features, match_features

# Define paths
image_directory = Path("path/to/your/images")
output_directory = Path("outputs/reconstruction")

# Define configurations
retrieval_config = extract_features.confs["netvlad"]
feature_config = extract_features.confs["superpoint_aachen"]
matcher_config = match_features.confs["superpoint+lightglue"]

# Create and run the pipeline
pipeline = ReconstructionPipeline(
    retrieval_conf=retrieval_config,
    feature_conf=feature_config,
    matcher_conf=matcher_config,
    num_matched_pairs=5,
)
pipeline.run(image_directory, output_directory)

Project Structure

The project is organized as follows:

├── pyproject.toml
├── README.md
├── run_pipeline.py
├── src
│   └── easy_3dgs
│       ├── __init__.py
│       ├── pipeline
│       │   ├── __init__.py
│       │   ├── orchestrator.py
│       │   ├── feature_extraction
│       │   ├── feature_matching
│       │   ├── feature_retrieval
│       │   ├── image_undistortion
│       │   ├── pair_generation
│       │   └── reconstruction
│       └── third_party
│           └── Hierarchical-Localization
└── test.py

• run_pipeline.py: An example script demonstrating how to use the ReconstructionPipeline.
• src/easy_3dgs/pipeline: Contains the core logic for the reconstruction pipeline.
• src/easy_3dgs/pipeline/orchestrator.py: The main entry point for the library, containing the ReconstructionPipeline class.
• src/easy_3dgs/third_party: Contains third-party libraries, such as Hierarchical Localization.

To Do

• Add 3d Gaussian Splatting

Contributing

Contributions are welcome! Please feel free to submit a pull request or open an issue.

License

This project is licensed under the MIT License. See the LICENSE file for details.