Improving the Performance of the Contextual Spaces Re-Ranking Algorithm on Heterogeneous Systems
Flávia Pisani, Daniel Carlos Guimarães Pedronette, Ricardo Torres, Edson Borin · 2015
Re-ranking algorithms have been proposed to improve the effectiveness of Content-Based Image Retrieval (CBIR) systems by exploiting contextual information encoded in distance measures and ranked lists. In this paper, we show how we improved the efficiency of one of these algorithms, called Contextual Spaces Re-Ranking. We propose a modification to the algorithm that reduces its execution time by 1.6x on average and improves its accuracy in most of our test cases. We also parallelized the implementation with OpenCL to use the CPU and GPU of an Accelerated Processing Unit (APU). Employing these devices to run different parts of the code resulted in speedups that range from 3.3x to 4.2x in comparison with the total execution time of the serial version.