3DGS Edit Tools

A Python library to convert 3D Gaussian Splatting (3DGS) format data to CSV format, edit it, and convert it back to 3DGS format.

Features

• Convert 3DGS format (PLY) data to CSV format
• Convert CSV format data to 3DGS format (PLY)
• Convert 3DGS format to standard point cloud format
• Convert point cloud data to/from CSV format for easy editing
• Tools for comparing 3DGS files and analyzing differences
• Merge multiple 3DGS files into a single scene
• Apply transformations (translation, scaling) to 3DGS files
• Verified compatibility with SuperSplat for visualization
• Point cloud data verified with CloudCompare

Project Structure

• src/: Core functionality for converting between formats
• tools/: Additional utilities for analyzing and comparing 3DGS files
• examples/: Sample code demonstrating usage of the library
• images/: Sample images and visualizations

Dependencies

Core Library (src/)

The core library in the src/ folder requires numpy for numerical operations. This dependency is automatically installed when you install the package.

Tools (tools/)

The tools in the tools/ folder require additional dependencies:

pip install pandas matplotlib

These are automatically installed if you install the package with the [tools] extra.

Installation

From PyPI (recommended)

# Basic installation with core functionality
pip install 3dgs-edit-tools

# Complete installation with tools for analysis and comparison
pip install 3dgs-edit-tools[tools]

From Source

Clone this repository:

git clone https://github.com/404background/3dgs-edit-tools.git
cd 3dgs-edit-tools
pip install -e .  # Install in development mode

Why CSV Format?

This library uses CSV format as an intermediate step for editing 3D Gaussian Splatting data because:

• Human-Readable: CSV files can be easily viewed and edited with spreadsheet software or text editors
• Ease of Editing: Unlike binary formats, CSV files allow direct modification of values without specialized tools
• Compatibility: CSV is widely supported by data processing tools and programming languages
• Transparency: The structure and values are visible, making it easier to understand the data

Point cloud data is also converted to CSV format before editing to maintain the same benefits.

Usage

As a Library

from src import convert_3dgs_to_csv, convert_csv_to_3dgs

# Convert PLY file to CSV
csv_path, _ = convert_3dgs_to_csv('model.ply')

# Write your code to edit the CSV file here
# ...

# Convert edited CSV back to PLY
restored_ply = convert_csv_to_3dgs(csv_path, None)

Merging and Transforming 3DGS Files

from src import merge_3dgs_files, create_transformed_copy

# Create a transformed copy with 10cm translation on X-axis
transformed_ply = create_transformed_copy('model.ply', 'model_moved.ply', 
                                         {'translate': [0.1, 0, 0]})

# Merge original and transformed models
merged_ply = merge_3dgs_files('model.ply', transformed_ply, 'merged_scene.ply')

Point Cloud Workflows

from src import (
    convert_3dgs_to_pointcloud, 
    convert_pointcloud_to_3dgs,
    convert_pointcloud_to_csv,
    convert_csv_to_pointcloud
)

# 3DGS to point cloud
pointcloud_path = convert_3dgs_to_pointcloud('model.ply')

# Point cloud to CSV for easy editing
csv_path = convert_pointcloud_to_csv(pointcloud_path)

# Edit CSV file here...

# CSV back to point cloud
edited_pointcloud = convert_csv_to_pointcloud(csv_path)

# Point cloud back to 3DGS
restored_model = convert_pointcloud_to_3dgs(edited_pointcloud, 'model.ply')

Using Tools

To use the comparison tool for analyzing differences between 3DGS files:

# First install required dependencies
pip install pandas matplotlib

# Run the comparison tool
python -m tools.compare_gs path/to/original.ply path/to/modified.ply --output-dir comparison_results

For more information about available tools, see the tools/README.md file.

From Command Line

After installation, the package provides several command-line executables for easy use:

Convert PLY to CSV:

3dgs-to-csv input.ply --output_csv output.csv

Convert CSV to PLY:

csv-to-3dgs input.csv --output_ply output.ply

Convert 3DGS to point cloud:

3dgs-to-pointcloud input.ply --output_ply output_pointcloud.ply

Convert point cloud to 3DGS:

pointcloud-to-3dgs input_pointcloud.ply original.ply --output_ply restored.ply

Merge two 3DGS files:

merge-gs file1.ply file2.ply --output merged.ply --translate-x 0.1

Convert 3DGS to mesh:

3dgs-to-mesh input.ply --output output_mesh.obj --method hybrid --quality high

Convert point cloud to CSV:

pointcloud-to-csv ply2csv input_pointcloud.ply --output_csv output.csv

Convert CSV to point cloud:

csv-to-pointcloud input.csv --output_ply output.ply --color_type uchar

Compare two 3DGS files:

compare-gs original.ply modified.ply --output-dir comparison_results

Sample Code

Sample code is available in the examples folder. See examples/README.md for details.

Visualization Tools

This project has been tested with the following visualization tools:

• SuperSplat: For 3D Gaussian Splatting format files (.ply)
• CloudCompare: For point cloud format files (.ply)
• Blender: For mesh format files (.obj, .ply, .stl)

All point cloud data conversions have been extensively tested using CloudCompare to ensure proper structure, color representation, and export compatibility. Mesh outputs have been verified with Blender to confirm proper mesh topology and compatibility.

API Reference

Core Conversion Functions

convert_3dgs_to_csv(ply_filename, csv_filename=None)

Converts 3DGS format (PLY) data to CSV format.

Arguments:

• ply_filename (str): Path to the input PLY file
• csv_filename (str, optional): Path to the output CSV file. If not specified, it's automatically generated from the input filename

Returns:

• tuple: (csv_filename, None) - Paths of the generated files

convert_csv_to_3dgs(csv_filename, footer_filename=None, output_ply_filename=None)

Converts CSV format data to 3DGS format (PLY).

Arguments:

• csv_filename (str): Path to the input CSV file
• footer_filename (str, optional): Path to the file containing footer data. If not specified, it's automatically generated from the input filename
• output_ply_filename (str, optional): Path to the output PLY file. If not specified, it's automatically generated from the input filename

Returns:

• str: Path of the generated PLY file

Merge and Transform Functions

merge_3dgs_files(file1, file2, output_file=None, transform=None)

Merges two 3D Gaussian Splatting files into a single file.

Arguments:

• file1 (str): Path to the first 3DGS file
• file2 (str): Path to the second 3DGS file
• output_file (str, optional): Path to the output merged 3DGS file
• transform (dict, optional): Transformation to apply to the second file (e.g. {'translate': [0.1, 0, 0]})

Returns:

• str: Path of the generated merged 3DGS file

create_transformed_copy(input_ply, output_ply, transformation)

Creates a transformed copy of a 3DGS file with specified transformations.

Arguments:

• input_ply (str): Path to input PLY file
• output_ply (str): Path to output transformed PLY file
• transformation (dict): Dictionary with transformation parameters (e.g. {'translate': [0.1, 0, 0], 'scale': [1.5, 1, 1]})

Returns:

• str: Path to transformed PLY file

Point Cloud Conversion Functions

convert_3dgs_to_pointcloud(ply_filename, output_ply_filename=None)

Converts 3D Gaussian Splatting format to standard point cloud PLY format.

Arguments:

• ply_filename (str): Path to the input 3DGS PLY file
• output_ply_filename (str, optional): Path to the output point cloud PLY file

Returns:

• str: Path of the generated point cloud file

convert_pointcloud_to_3dgs(pointcloud_ply, original_3dgs_ply, output_ply=None)

Converts a point cloud file back to 3D Gaussian Splatting format using an original 3DGS file as reference.

Arguments:

• pointcloud_ply (str): Path to the input point cloud PLY file
• original_3dgs_ply (str): Path to the original 3DGS file to use as reference
• output_ply (str, optional): Path to the output 3DGS file

Returns:

• str: Path of the generated 3DGS file

convert_pointcloud_to_csv(ply_filename, csv_filename=None)

Converts point cloud PLY file to CSV format for easier editing.

Arguments:

• ply_filename (str): Path to the input point cloud PLY file
• csv_filename (str, optional): Path to the output CSV file

Returns:

• str: Path of the generated CSV file

convert_csv_to_pointcloud(csv_filename, output_ply_filename=None, color_type="uchar")

Converts CSV file to point cloud PLY format.

Arguments:

• csv_filename (str): Path to the input CSV file
• output_ply_filename (str, optional): Path to the output point cloud PLY file
• color_type (str): Type of color data to use in PLY file ('float' or 'uchar')

Returns:

• str: Path of the generated PLY file

License

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