Cost-Effective HPC Clustering for Computer Vision Applications

We will present a cost-effective and flexible realization of high performance computing (HPC) clustering and its potential in solving computationally intensive problems in computer vision. The featured software foundation to support the parallel programming is the GNU parallel Knoppix package with message passing interface (MPI) based Octave, Python and C interface capabilities. The implementation is especially of interest in applications where the main objective is to reuse the existing hardware infrastructure and to maintain the overall budget cost. We will present the benchmark results and compare and contrast the performances of Octave and MATLAB.