Abstract—In our ongoing research we focus on detection of firearms and knives in video stream from Closed Circuit Video Television. During this research we came across a multiple-objective optimization problem. Our system operation depends on a set of launch parameters that differ between experiments. In this paper we describe and compare several automated optimization algorithms used for selection of these parameters. We apply those algorithms in order to select the best configuration parameters and compare them in the context of our system aiming at automated detection of dangerous tools.
Index Terms—hyperparameter optimization, genetic algorithms, random search, Bayesian optimization, simulated annealing, firearms detection
Cite: Jakub Król and Michał Grega, "Optimization of Configuration Parameter Set in Video Analysis," Journal of Image and Graphics, Vol. 7, No. 4, pp. 130-133, December 2019. doi: 10.18178/joig.7.4.130-133
Copyright © 2012-2020 Journal of Image and Graphics, All Rights Reserved