OpenMVG

OpenMVG is a C++ library and pipeline for reconstructing 3D scenes from images with multiple-view geometry and structure-from-motion methods. It suits developers and researchers building or studying image-based reconstruction workflows. Its open-source tools cover much of the route from image features to scene export, while dense point clouds and textured meshes depend on other software.

Overview

OpenMVG combines libraries, binaries, and pipelines for 3D reconstruction. The libraries cover image manipulation, feature description and matching, feature tracking, camera models, multiple-view geometry, robust estimation, and structure-from-motion algorithms. Binaries handle scene initialization, feature detection and matching, reconstruction, and export to other multiple-view-stereo frameworks.

That breadth makes OpenMVG a flexible foundation for building or adapting a reconstruction workflow. It is a better fit for people who want to work with algorithms and pipelines than for those seeking a finished capture app. The project emphasizes readable implementations of classic and state-of-the-art methods to support reproducible research, and describes the software as designed to be read, learned from, modified, and used. Test-driven development and samples support that aim.

Key features

Reconstruction and export

OpenMVG carries an image dataset through scene initialization, feature processing, and structure-from-motion reconstruction. It exports reconstructed scenes for use with other multiple-view-stereo frameworks, which can compute dense point clouds or textured meshes. This division is useful when a team wants control over the reconstruction stage, but it means those dense outputs rely on another framework.

Libraries and learning resources

The library spans feature tracking, camera models, robust estimation, and multiple-view geometry as well as image and feature processing. Tutorials cover data structures, library use, and processing an image dataset; samples and the wiki provide further usage resources. That makes the project relevant both as a component for custom pipelines and as a codebase for studying reconstruction methods.

Ground control point support and API access add options for workflows that need those capabilities. The project is developed in C++ and is licensed under Mozilla Public License 2.0, so it is aimed at users comfortable working with a software library and its binaries.

Pricing

OpenMVG costs nothing. The OpenMVG plan includes an open-source library under the MPL-2.0 license, with API access, point cloud export, and ground control point support. There is no paid tier to move to for the listed capabilities, which suits research groups and developers seeking a free base for custom work.

Platforms

OpenMVG is listed for Android, iOS, Linux, macOS, self-hosted use, and Windows, with desktop deployment. Its framework approach and C++ implementation make it most suitable for technical users building a workflow around the library and binaries.

Who it's for

OpenMVG is a strong choice for developers and researchers who need to inspect, modify, or assemble image-based 3D reconstruction methods. Reproducibility, readable implementations, samples, and tutorials serve that audience. It is less suitable for someone who wants a complete scanning experience or expects this tool alone to produce dense point clouds and textured meshes.

Pros and cons

• Broad reconstruction toolkit: libraries and binaries span image handling through structure-from-motion and scene export, giving technical users multiple stages to work with.
• Research-friendly codebase: readable implementations, samples, and test-driven development support learning and modification.
• Free and open source: the plan costs nothing and uses the MPL-2.0 license.
• Dependent dense outputs: dense point clouds and textured meshes are computed with other multiple-view-stereo frameworks, adding another tool to the workflow.
• Technical orientation: a C++ library and binaries suit people building software workflows better than users looking for a finished capture app.

Alternatives

COLMAP is another free option, with Linux, macOS, web, and Windows platforms; consider it when those platform choices fit your workflow.

MicMac is a free, open-source photogrammetric suite under the CeCILL-B license, available for Linux, macOS, and Windows. Choose it when a photogrammetric suite is a closer fit.

PIX4Dcatch has a free Discovery plan with LiDAR 3D capture, GPS georeferencing, point measurement, and GCP creation and marks. Its Android, iOS, and Windows support and capture-focused plan make it a candidate for users who need those functions.

Polycam offers a free plan with limited space, object and AI captures, GLTF exports, and public link sharing, across Android, iOS, and web. It may suit users looking for a capture and sharing workflow.

RealityScan is a free option for individuals and small businesses under $1 million USD in revenue in the past 12 months, educational institutions, and students, with Android, iOS, Linux, and Windows support.

WebODM is free to install and use, runs locally and offline, and supports API, Linux, macOS, self-hosted, web, and Windows deployments. Consider it when local offline operation or those deployment choices matter.

KIRI Engine has a free Basic plan for up to 150 photos per scan, a 2 GB scan limit, unlimited scans, and free exports, across Android, API, iOS, and web. It is a candidate when those scan limits and platforms fit.

MetaScan has a free plan with up to five scans in its library and one HD reconstruction per month, alongside full-quality preview scans, sharing, and figurine ordering, for Android, iOS, and web.

For more options, browse Photogrammetry Software or AI 3D Scanning Software.

Verdict

Choose OpenMVG if you are a developer or researcher who wants a free, readable, modifiable foundation for image-based 3D reconstruction, with API access, ground control point support, and scene export. Look elsewhere if you need a finished scanning app or require dense point clouds and textured meshes without relying on another framework.