Autodesk’s 123D Catch is great for making 3D scans from digital photos — miraculous, really — but it’s a cloud-based freebie that comes with constraints. You can upload only a limited number of photos, and your 3D mesh will come back at a maximum resolution that may not be all you desire. Here are three ways to create your own 3D models with photogrammetry, aka “structure from motion (SFM)”, using free open source software:
VisualSFM — Digital compositor Jesse Spielman wondered if he could match 123D Catch’s results with open source tools, spending zero dollars. A total noob, he dove in and a few weeks later wrote this incredible blog post showing how to use VisualSFM to roll your own point clouds without arbitrary limits on the number of photos or point cloud density, then use Meshlab to generate a mesh and paint it with the full-resolution photo texture. [Props to Matt Griffin at Adafruit for tipping us.] You’re only limited by your computing power. (Devil’s advocate: Is that why we’re using 123D Catch in the first place?)
VisualSFM was developed by Changchang Wu at the University of Washington as an outgrowth of the open-source frameworks Bundler (to do multi-view calibration and create sparse point clouds), CMVS/PMVS (to compute dense 3D point clouds), and SIFT. Download it here. Changchang is now working on 3D computer vision for Google, BTW.
Python Photogrammetry Toolbox — This set of Python scripts automates Bundler and CMVS/PMVS, fronted by an open source GUI. Archaeologists are using it for handheld and aerial photo reconstruction of sites, remains, and artifacts. Download it here and check out Arc-Team’s recent tutorial here.
ArcheOS — An entire operating system built around Debian Squeeze to connect open photogrammetry and 3D modeling programs without the hassle of installing all the dependencies. ArcheOS wraps up the PPT and PPT GUI with Meshlab and more. Check it out here.