Skip to content
Switch branches/tags

Latest commit


Git stats


Failed to load latest commit information.
Latest commit message
Commit time

3D Photogrammetry Reconstruction Example

Example scripts and data from my Bristech talk. Slides at


The below are tested on Ubuntu 16.04, but will also work in a Cygwin environment with Windows MVE binaries and appropriate changes to the scripts (just point them at the location of the binaries).

It works on a Mac too (tested on High Sierra version 10.13.3) if you follow these build instructions..

Clone and build MVE in this directory:

git clone
cd mve
make -j

If you want to use UVME to visualise the steps then also:

cd mve
cd apps/umve
qmake -qt=qt5
make -j

See the MVE repo for the prerequisites for it.

I want to make a 3D model!

Put all your images in a folder, then run scripts 0-5 in turn with that name as an argument.

Provided as a test/example are my Engineshed photos. using that as an example:

./ engineShed

You'll now have a engineShed_scene folder, you can open this scene in UVME and see the photos. This is where the output goes.

./ engineShed

You'll run the sfm, have sparse features and camera locations. Reload the scene and look at the scene inspector. Check "draw sfm points" - you might need to move the view.

./ engineShed

You now have per-image depth maps.

./ engineShed

You can load pset-L2.ply in the scene viewer, this is your dense cloud.

./ engineShed

Now you have surface-L2.ply to load - the surface reconstruction.

./ engineShed

Now you've got the final cleaned mesh - surface-L2-clean.ply. It should look as below:

alt text


Example Scripts from my Bristech talk on 3D reconstruction April 2018



No releases published


No packages published