No abstract

In this paper we investigate the application of the Particle Swarm Optimization (PSO) technique for solving the Hardware/Software partitioning problem. The PSO is attractive for the Hardware/Software partitioning problem as it offers reasonable coverage of the design space together with O(n) main loop's execution time, where n is the number of proposed solutions that will evolve to provide the final solution. We carried out several tests on a hypothetical, relatively-large Hardware/Software partitioning problem using the PSO algorithm as well as the Genetic Algorithm (GA), which is another evolutionary technique. We found that PSO outperforms GA in the cost function and the execution time. For the case of unconstrained design problem, we tested several hybrid combinations of PSO and GA algorithms; including PSO then GA, GA then PSO, GA followed by GA, and finally PSO followed by PSO. We found that a PSO followed by GA algorithm gives small or no improvement at all, while a GA then PSO algorithm gives the same results as the PSO alone. The PSO algorithm followed by another PSO round gave the best result as it allows another round of domain exploration. The second PSO round assign new randomized velocities to the particles, while keeping best particle positions obtained in the first round. We propose to name this successive PSO algorithm as the Re-excited PSO algorithm.

No abstract

No abstract

Object trackers based on Convolution Neural Network (CNN) have achieved state-of-the-art performance on recent tracking benchmarks, while they suffer from slow computational speed. The high computational load arises from the extraction of the feature maps of the candidate and training patches in every video frame. The candidate and training patches are typically placed randomly around the previous target location and the estimated target location respectively. In this paper, we propose novel schemes to speedup the processing of the CNN-based trackers. We input the whole region-of-interest once to the CNN to eliminate the redundant computations of the random candidate patches. In addition to classifying each candidate patch as an object or background, we adapt the CNN to classify the target location inside the object patches as a coarse localization step, and we employ bilinear interpolation for the CNN feature maps as a fine localization step. Moreover, bilinear interpolation is exploited to generate CNN feature maps of the training patches without actually forwarding the training patches through the network which achieves a significant reduction of the required computations. Our tracker does not rely on offline video training. It achieves competitive performance results on the OTB benchmark with 8x speed improvements compared to the equivalent tracker.