v3p - Vuer 3DGS Processing

Tools for Gaussian Splatting (3DGS) processing workflows. Built with params-proto for configuration management.

Installation

pip install v3p

For development:

git clone https://github.com/vuer-ai/vuer-3dgs-processing.git
cd vuer-3dgs-processing
pip install -e ".[dev]"

Quick Start

CLI Usage

The v3p command provides several subcommands for processing 3DGS point clouds:

# Show available commands
v3p

# Process a point cloud
v3p process --input-path input.ply --output-path output.ply

# Convert between formats
v3p convert --input-path input.ply --output-path output.pcd

Python API

import v3p
print(v3p.__version__)

Using the configuration classes directly:

from v3p.scripts.process import ProcessConfig, main

# Configure via class attributes
ProcessConfig.input_path = "input.ply"
ProcessConfig.output_path = "output.ply"
ProcessConfig.voxel_size = 0.01
ProcessConfig.verbose = True

# Run processing
main()

Commands

v3p process

Process 3DGS point clouds with filtering and downsampling.

v3p process --input-path <path> [options]

Options:

• --input-path: Path to input PLY or point cloud file (required)
• --output-path: Path to output file (default: output.ply)
• --voxel-size: Voxel size for downsampling (default: 0.01)
• --remove-outliers: Remove statistical outliers (default: True)
• --num-neighbors: Number of neighbors for outlier detection (default: 20)
• --std-ratio: Standard deviation ratio for outlier removal (default: 2.0)
• --verbose: Enable verbose output

v3p convert

Convert between point cloud formats.

v3p convert --input-path <path> --output-path <path> [options]

Options:

• --input-path: Path to input file (required)
• --output-path: Path to output file (required)
• --input-format: Input format - auto, ply, pcd, xyz (default: auto)
• --output-format: Output format - ply, pcd, xyz (default: ply)
• --binary: Write binary format when supported (default: True)
• --verbose: Enable verbose output

Configuration with params-proto

v3p uses params-proto for configuration management. This allows you to:

1. Set parameters via CLI: All config class attributes become CLI flags
2. Set parameters programmatically: Modify class attributes directly
3. Use sweep files: Generate parameter sweeps for batch processing

Example sweep file:

from params_proto import Sweep
from v3p.scripts.process import ProcessConfig

with Sweep(ProcessConfig) as sweep:
    ProcessConfig.input_path = "input.ply"
    with sweep.product:
        ProcessConfig.voxel_size = [0.01, 0.02, 0.05]

sweep.save("process_sweep.jsonl")

Development

# Install dev dependencies
pip install -e ".[dev]"

# Run tests
make test

# Build and preview documentation
make preview

# Build wheel
make wheel

# Publish to PyPI
make publish

Documentation

Full documentation: https://v3p.readthedocs.io/en/latest/

License

MIT