Installing Decapod Stereo Dewarping Notes

Dewarp

Download the dewarp code here: https://bitbucket.org/mkraemer/decapod-dewarping

Installation Notes

DECAPOD STEREO DEWARPING
========================

required
------------------
- numpy
- scipy
- python-matplotlib
- git (to download pyflann from source)
- opencv
- pyflann
- build-essential

git

• Required for downloading pyflann from source.

numpy

• Note numpy should be installed prior to pyVLFeat
• Install sudo apt-get install python-numpy
• Version assuming v1.6.2?

scipy

• Install sudo apt-get install python-scipy
• Version assuming 0.11.0rc2?

python-setuptools

• Note Python Setuptools is required to install pyflann from source.
• Install sudo apt-get install python-setuptools
• Version ?

opencv

• follow the instructions here: http://www.samontab.com/web/2012/06/installing-opencv-2-4-1-ubuntu-12-04-lts/
  • don't have to continue after make install

pyflann

• URL http://www.cs.ubc.ca/~mariusm/index.php/FLANN/FLANN
• Install https://github.com/mariusmuja/flann/issues/15
• License BSD (see source code for license info)
• Note: Install according to link above. Copy pyflann directory from usr/local/lib/python2.7/dist-packages/ to ~/decapod-dewarping/

vlfeat

• URL http://www.vlfeat.org/index.html
• download binary or compile from source
• see http://www.vlfeat.org/install-shell.html for locations of files
  • copy "sift" to "/usr/local/bin
  • copy "libvl.so" to "/usr/local/lib
  • run ldconfig

example usage - calibration
---------------------------
./calibrate.py calibration-images 9 6 calibration-data

This uses the chessboard images in directory calibration-images. Each
chessboard has 9x6 inner corners. The resulting camera matrices are
persisted into directory calibration-data.

example usage - dewarping
-------------------------
./dewarp.py calibration-data left.jpg right.jpg dewarped.png

This uses the calibration matrices from directory calibration-data to
perform stereo dewarping of the image pair given by left.jpg and right.jpg.
The dewarped output is then persisted to dewarped.png

Calibration notes

• 6 minutes to run calibration on 2 cores, 2 GB VM.
• 6 minutes to run calibration on 4 cores, 2 GB VM.