Turn your vacation photos into 3D geometry, no lidar required

What it does AliceVision takes a pile of regular photos or video frames and works backwards from 2D to 3D: it figures out camera positions, builds a sparse point cloud, then densifies it into a mesh. It also handles camera tracking for matchmoving and can generate HDR panoramas. The output is production-grade geometry, not a toy demo.

The interesting bit This is academic research that survived contact with Hollywood. Born from a 2010 partnership between French VFX house Mikros Image and university labs, it merged multiple PhD theses and EU-funded projects into a single pipeline. The codebase carries that lineage openly: it maintains a full bibliography of papers it implements.

Key highlights

• Structure-from-motion, multiview stereo, meshing, and texturing in one pipeline
• Camera tracking with support for CCTag markers and natural features
• HDR panorama stitching from bracketed exposures
• MPLv2 license, not GPL, so commercial derivatives face fewer friction
• Sister project Meshroom provides both a Qt GUI and headless Python scripting

Caveats

• The README is thin on technical specifics; you will need to dig into the website and bibliography for algorithmic details
• Building from source is non-trivial enough that the project maintains a separate INSTALL.md
• The project history is complex; multiple merged codebases can mean uneven documentation

Verdict Grab this if you need open-source photogrammetry with proven production credentials—especially for VFX, cultural heritage digitization, or surveying. Skip it if you want a one-click mobile app; the learning curve is real and the tooling assumes you know why parallax matters.