Ciencias Exactas y Ciencias de la Salud

Now showing 1 - 2 of 2

Detection of suspicious attitudes on video using neuroevolved shallow and deep neural networks models
(Instituto Tecnológico y de Estudios Superiores de Monterrey, 2021-11) Flores Munguía, Carlos; Terashima Marín, Hugo; puemcuervo/tolmquevedo; Oliva, Diego; Ortiz Bayliss, Jose Carlos; School of Engineering and Sciences; Campus Monterrey
The analysis of surveillance cameras is a critical task usually limited by the people involved in the video supervision devoted to such a task, their knowledge, and their judgment. Security guards protect other people from different events that can compromise their security, like robbery, extortion, fraud, vehicle theft, and more, converting them to an essential part of this type of protection system. If they are not paying attention, crimes may be overlooked. Nonetheless, different approaches have arisen to automate this task. The methods are mainly based on machine learning and benefit from developing neural networks that extract underlying information from input videos. However, despite how competent those networks have proved to be, developers must face the challenging task of defining the architecture and hyperparameters that allow the network to work adequately and optimize the use of computational resources. Furthermore, selecting the architecture and hyperparameters may significantly impact the neural networks’ performance if it is not carried out adequately. No matter the type of neural network used, shallow, dense, convolutional, 3D convolutional, or recurrent; hyperparameter selection must be performed using empirical knowledge thanks to the expertise of the designer, or even with the help of automated approaches like Random Search or Bayesian Optimization. However, such methods suffer from problems like not covering the solution space well, especially if the space is made up of large dimensions. Alternatively, the requirement to evaluate the models many times to get more information about the evaluation of the objective function, employing a diverse set of hyperparameters. This work proposes a model that generates, through a genetic algorithm, neural networks for behavior classification within videos. The application of genetic algorithms allows the exploration in the hyperparameters solution space in different directions simultaneously. Two types of neural networks are evolved as part of the thesis work: shallow and deep networks, the latter based on dense layers and 3D convolutions. Each sort of network takes distinct input data types: the evolution of people’s pose and videos’ sequences, respectively. Shallow neural networks are generated by NeuroEvolution of Augmented Topologies (NEAT), while CoDeepNEAT generates deep networks. NEAT uses a direct encoding, meaning that each node and connection in the network is directly represented in the chromosome. In contrast, CoDeepNEAT uses indirect encoding, making use of cooperative coevolution of blueprints and modules. This work trains networks and tests them using the Kranok-NV dataset, which exhibited better results than their competitors on various standard metrics.
Jones Matrix Characterization of Homogeneous Optical Elements via Evolutionary Algorithms
(Instituto Tecnológico y de Estudios Superiores de Monterrey, 2021-06-15) De Luna Pámanes, Alejandra; COVANTES OSUNA, EDGAR; 352304; Covantes Osuna, Edgar; tolmquevedo/mscuervo; Amaya Contreras, Iván Mauricio; Ortiz Bayliss, José Carlos; Serrano García, David Ignacio; School of Engineering and Sciences; Campus Monterrey; López Mago, Dorilián
Jones calculus provides a robust and straightforward method to describe polarized light and polarizing optical systems using two-element vectors (Jones vectors) and 2 X 2 matrices (Jones matrices). Jones matrices are used to determine the retardance and diattenuation introduced by an optical element or a sequence of elements. Moreover, they are the tool of choice to study optical geometric phases, the polarization-dependent phase of the total delay of a light beam acquired when passing through a material. Jones matrix characterization is a technique used to characterize polarizing optical systems. By measuring the geometric phase, Jones matrix characterization can identify the sample's eigenpolarizations, which are those polarization states that exits the sample only scaled by a phase factor. Currently, there is only one existing Jones matrix characterization method available. However, said method is inefficient, since the characterization of any given element is time-consuming given that the method is based on a general sampling strategy. Optimization techniques are used to find a solution to a problem specified by an objective function, where the variables are searched over to find the combination that results in the best objective function value while satisfying the constraints of the problem. Evolutionary Algorithms (EAs) are optimization techniques based on the theory of evolution, which explains the adaptive changes of species in nature through the survival of the fittest, heredity, and mutation. They are all random-based meta-heuristic algorithms that do not require gradient information and typically make use of several points in the search space at a time. Therefore, using the exploration capabilities of EAs, in this study, we present an initial approach for solving the problem of finding the eigenvectors that characterize the Jones matrix of a homogeneous optical element through EAs. We evaluate the analytical performance of an EA with a polynomial mutation (PM) operator and a Genetic Algorithm (GA) with a simulated binary crossover operator and a PM operator, and compare the results with those obtained through a general sampling method. The results show that both the EA and the GA out-performed a general sampling method of 6,000 measurements, by requiring in average 103 and 188 fitness functions measurements respectively, while having a perfect rate of convergence. The present analysis shows that the usage of EAs in the area of optics is a promising research area and as future research, we would like to apply EAs on the more complex case of inhomogeneous optical elements, for which no method of characterization currently exists.

Ciencias Exactas y Ciencias de la Salud

Browse

Filters

Settings

Sort By

Results per page

Search Results