yeicor/colmap-openmvs repository overview

⁠colmap-openmvs

A robust, production-ready photogrammetry pipeline combining COLMAP⁠ and OpenMVS⁠, containerized for reproducibility and ease of use. Supports both CPU-only and CUDA-accelerated environments: run the full pipeline on any machine, or leverage GPU acceleration for faster processing when available.

⁠Features

• End-to-End Photogrammetry: From unordered images to textured 3D meshes, fully automated.
• CPU & CUDA Support: Choose CPU for maximum compatibility (even on GPU-less servers), or CUDA for optimal speed (with a larger image size). Both are first-class citizens.
• Latest Development Versions: Always up-to-date with automated, tested builds.
• Dockerized: Run anywhere with a single command; no dependency hell.
• Intelligent Caching: Only re-runs steps when inputs change.
• Verbose Logging & Dry-Run: Debug and inspect every step, or simulate runs without execution.
• User Mapping: Use -u $(id -u):$(id -g) for proper file permissions on output.
• Comprehensive Help: --help provides detailed info about all embedded tools and configuration options.

⁠Quickstart

1. Prepare your images

Place your images in a directory with a subfolder called images/:

/your/data/path/
  images/
    img1.jpg
    img2.jpg
    ...

1. (Optional) Add a custom config

You can create a config.sh file in your data directory to override environment variables and fine-tune the pipeline.

To see all available options and configuration variables:

docker run --rm -it yeicor/colmap-openmvs:cpu-latest --help

1. Run the pipeline using Docker

For CPU-only (compatible with all systems):

docker run --rm -it -u $(id -u):$(id -g) \
  -v /your/data/path:/data \
  yeicor/colmap-openmvs:cpu-latest /data

For CUDA acceleration (requires NVIDIA GPU and drivers):

docker run --rm -it --gpus all -u $(id -u):$(id -g) \
  -v /your/data/path:/data \
  yeicor/colmap-openmvs:cuda-latest /data

Happy reconstructing!