Projects
Gaussian Processes for Machine Learning
|
In recent years Gaussian Processes have become more and more popular for doing machine learning. A Gaussian Process can be seen as an infinite dimensional Gaussian Distribution defined by a mean function and a covariance function. Using Gaussian Processes for non-linear regression only involves the choice of such a covariance function (the mean function is often neglected due to missing knowledge) which could be defined by a radial basis function with some learnable hyperparameters. Here you find more information on how Gaussian Processes can be used to do Machine Learning including a Java Applet showing Gaussian Process Regression and the influence of hyperparameters. |
Automatic Multiple Camera Calibration
  |
At EPFL we developed an algorithm in order to calibrate multiple cameras at once with a planar calibration object. Feature detection is done by an algorithm of Lepetit. Initial guessing ist done based on a method proposed by Zhang. Afterwards global optimization is processed to refine intrinsic and extrinsic camera parameters. Additionally discrete light sources position, direction and strength is estimated. This allows for realistic augmented reality rendering. The whole process only takes 5 minutes. You will find more information at the site of CVLAB. You can try it yourself with your webcam by downloading the sources or binaries. Compiling the binaries requires the free OpenCV library.
|
Linux/Windows Binaries and Source code (English)
