Meshroom developers have released Meshroom 2019.1, the first major open source software update for photogrammetry since its initial release.

The upgrade improves the speed and quality of mesh generation, adds support for source images captured using projected light patterns, and eliminates the need to use an Nvidia GPU.

Based on a powerful open source photogrammetry framework

Officiallybreleased last year, Meshroom is a graphical interface for AliceVision, the well-known open source photogrammetry platform.

The framework generates textured meshes from sets of photos of real objects, exporting geometry in OBJ format. Intermediate point cloud data can be exported in Alembic format, and depth maps as EXR files.

There’s no need in Nvidia

In Meshroom 2019.1, a “more accurate, faster, and customizable” estimate of the space to be restored was given based on the intermediate point cloud. The new algorithm supports high density areas in the cloud and removes poorly defined ones.

The update also allows for mashing on systems with video cards other than Nvidia, although using the Draft Meshing option on a lower quality. The standard method, which involves creating depth maps, is based on CUDA and requires the use of an Nvidia GPU.

Support for projected light patterns, better matching of real-world cameras

Meshroom 2019.1 also allows the restoring of meshes from images taken using projected light patterns using a separate set of images as textures.

In addition, the sensor database used by the software has been expanded to improve the match of real camera models. Also, it is now possible to display internal camera information extracted from metadata.

Availability and System Requirements

AliceVision and Meshroom are available for free under the MPL2 license. Compiled Meshroom binaries are available for 64-bit Windows and Linux. Although there is no official version of software for MacOS, you can find instructions for compiling it from the source code here.