Vagrant machine for using OpenDroneMap, meant primarily for Windows users who want to leverage ODM.
- Install VirtualBox ( http://virtualbox.org )
- Install Vagrant ( http://vagrantup.com )
- (For Windows and Mac) Install Github ( http://windows.github.com or http://mac.github.com )
- Make a directory c:\users\yourusername\Documents\odm\vagrant_data
- Copy 'odm_vagrant' to c:\users\yourusername\Documents\odm\
- Launch GitHub Command Line
- Navigate to vagrant directory.:
cd c:\users\yourusername\Documents\OpenDroneMap\
- Type the following in the command line
mkdir ..\vodm_data
vagrant up
vagrant ssh
- Clone zip of https://github.com/OpenDroneMap/OpenDroneMap.git and https://github.com/OpenDroneMap/odm_data.git
cd /vagrant_data/
sudo apt-get install git
git clone https://github.com/OpenDroneMap/OpenDroneMap.git
git clone https://github.com/OpenDroneMap/odm_data.git
- Install:
cd /vagrant_data/OpenDroneMap
./install.sh
Wait patiently... Now let us run OpenDroneMap on our test dataset.
cd /vagrant_data/odm_data/odm_data/
/vagrant_data/OpenDroneMap/./run.pl
Wait patiently again...
Outputs will be in /vagrant_data/odm_data/reconstruction-with-image-size-1200 and /vagrant_data/odm_data/reconstruction-with-image-size-1200-results .
In your Windows host, these will be at c:\users\yourusername\Documents\odm\vagrant_data\odm_data\reconstruction-with-image-size-1200
Install MeshLab 1.3.3 or later on your Windows host. Then...
From Meshlab 1.3.3:
* Open Project file, navigate to:
* c:\users\yourusername\Documents\odm\vagrant_data\odm_data\reconstruction-with-image-size-1200\reconstruction-with-image-size-1200\bundle\bundle.out
* It will prompt for the image list file
* c:\users\yourusername\Documents\odm\vagrant_data\odm_data\reconstruction-with-image-size-1200\reconstruction-with-image-size-1200\list.txt
* Control-L and delete "0 model"
* Import dense point cloud:
* e.g. c:\users\yourusername\Documents\odm\vagrant_data\odm_data\reconstruction-with-image-size-1200\reconstruction-with-image-size-1200-results\option-0000.ply
* (there may be multiple ply files)
* Make a mesh:
* Filters:Remeshing, Simplification and Reconstruction:Surface Reconstruction Poisson
* Texture the mesh
* Parameterization + texturing from registered rasters
Celebrate. Now try your own dataset. If you don't have one, saunter over to http://www.publiclab.org/ for DIY balloon advice, or get yourself a nifty quad copter with a commodity camera (preferably not a GoPro for now until we have a routine to calibrate those). Aim for 70-80% overlap between your photos.