Toward a distributed benchmarking tool for biometrics

Research in computer science evolves very quickly. In order to prove the efficiency of a new algorithm, it is generally necessary to show some results on a large and significant benchmark used in the state of the art. This fact has for consequence the need of a large computation capability in a research laboratory. We address in this paper the performance evaluation of biometric systems through distributed computing. In this domain, we need to realize with the same algorithm many computations on different data (generally corresponding to images). We propose a solution based on a software we developed for facilitating this task. The proposed architecture is composed of a server distributing computation tasks on all the available clients. Experimental results show the benefit of the proposed software.