Fast kNN search using GPU
|
The k-nearest neighbor (kNN) search is a problem found in many research and industrial domains such as 3-dimensional object rendering, content-based image retrieval, statistics (estimation of entropies and divergences), biology (gene classification), etc.
We propose two GPGPU implementations of the brute-force (exhaustive) kNN search algorithm. These implementations, through the APIs NVIDIA CUDA and NVIDIA CUBLAS, paralellize the kNN search process on the computer's graphic card.
|
jMEF - Mixtures of exponential families
|
jMEF is an open source Java framework to create, process and manage mixtures of exponential family.
An exponential family is a generic set of probability distributions that including most of the commonly used distributions such as Gaussian, Poisson, Bernoulli, binomial, multinomial, Laplacian, Gamma, etc.
Mixtures of exponential families provide a generic framework for handling Gaussian mixture models, mixture of Poisson distributions, and Laplacian mixture models as well.
|
RPI - Registration Programming Interface
|
RPI is a C++ framework for image registration providing a simple and intuitive interface covering most of image registration problems.
RPI uses ITK image and transformation objects: the standard I/O procedures are directly handled by ITK which is developed and maintained by hundred of developers.
RPI is however not based on the ITK registration framework considered as too strict and not generic enough.
RPI is simple, flexible and generic. It allows users to implement and integrate any kind of registration method, from the outdated C of fortran codes to the recent GPU implementations.
It simplifies the user experience while helping the author of a new registration method to focus on the algorithm, not on the user interface.
|
Keypoint extraction
|
Matlab implementation of some of the most common keypoint extraction algorithms:
|