Some projects are shared on github: my github account
As part of my PhD project I develop a markerless augmented reality framework.
Here I try to share what I learn, briefly. And this is by no means a complete guide.
Hopefully I will have time to complete these short articles.
2. Irrlicht engine for augmented reality
3. Feature tracking
4. Fundamental matrix, epipolar geometry
5. Camera calibration, essential matrix
Linux driver for Sony PS3 Eye camera
For my projects I use a Sony PS3 Eye camera. Currently the Linux driver, gspca_ov534, of this inexpensive camera supports very high frame rates, 60 FPS at VGA and 120 FPS at QVGA. I added a power-line frequency control to the driver to remove light flicker in the PAL region. This patch has been added to the http://linuxtv.org/hg/v4l-dvb master development tree.
To evaluate the performance of image matching algorithms (for location recognition), I needed to collect a large set of images along with their latitude and longitude coordinates while walking on street. To have a fast and reliable system for collecting large datasets I decided to connect a GPS and webcam to a laptop, and after reading the coordinates from GPS at 1 Hz, capture a frame and save. I used an Acer Aspire One with UNR installed on it, gpsd 2.92, SQLite, a PS3 Eye camera, and a Holux m-241. The code is copied at http://www.cs.bris.ac.uk/~ebrahimi/code/geotag/.
SUSurE feature detector
Update August 2010: the optimized implementation is ready, please contact me if you want the optimized code.
In 2009 we designed a fast scale-invariant feature detector, SUSurE. The paper was presented in "Workshop on Feature Detectors and Descriptors: The State Of The Art and Beyond", one of the CVPR09 workshops. This tarball contains a sample implementation of the algorithm (released under GPL 3). This implementation uses LodePNG to read PNG files and Cairo is used to mark the detect features. Please cite the paper, if you use this code for your research work.
Kinect point cloud viewer
Angkor is an open source project using Open Scene Graph and AntTweakBar libraries.