Large Scale 3D Shape Retrieval Based on Multi-core Architectures.

Despite of the variety of approaches proposed in the literature in order to improve the execution time of the 3D shape retrieval [14,15], the challenge that still remains is to design a 3D shape retrieval method that allows the large scale retrieval and, in the same time, respects the relevance of the obtained results. In this work, we deal with the problem of the large scale of 3D shape retrieval by proposing new implementations on multi-core environment. At our knowledge, a few partial works based on HPC High Performance Computing, have been proposed in the literature [1,2]. The proposed solutions are designed for the GPU Graphical Processing Unit and concern only the step of the extraction of the SIFT salient local features. In order to optimally exploit the potential of the multi-core architectures, we have studied different data distributions. Experimental results, under OpenMP environment, show that the large scale retrieval can be achieved using the multi-core environment.