Angélica González Arrieta

Deep neural networks and transfer learning applied to multimedia web mining

The growth in the amount of multimedia content available online supposes a challenge for search and recommender systems. This information in the form of visual elements is of great value to a variety of web mining tasks; however, the mining of these resources is a difficult task due to the complexity and variability of the images. In this paper, we propose applying a deep learning model to the problem of web categorization. In addition, we make use of a technique known as transfer or inductive learning to drastically reduce the computational cost of the training phase. Finally, we report experimental results on the effectiveness of the proposed method using different classification methods and features from various depths of the deep model.

A brief review of the ear recognition process using deep neural networks

Abstract The process of precisely recognize people by ears has been getting major attention in recent years. It represents an important step in the biometric research, especially as a complement to face recognition systems which have difficult in real conditions. This is due to the great variation in shapes, variable lighting conditions, and the changing profile shape which is a planar representation of a complex object. An ear recognition system involving a convolutional neural networks (CNN) is proposed to identify a person given an input image. The proposed method matches the performance of other traditional approaches when analyzed against clean photographs. However, the F1 metric of the results shows improvements in specificity of the recognition. We also present a technique for improving the speed of a CNN applied to large input images through the optimization of the sliding window approach.

Development of Sign Language Communication Skill on Children through Augmented Reality and the MuCy Model

This paper shows a Sign Language Teaching Model (SLTM) called: Multi-language Cycle for Sign Language Understanding (MuCy). It serves as complementary pedagogical resource for Sign Language (SL) teaching. A pilot lesson with the Rainbow Colors was conducted at the Association of Parents of Deaf Children of Salamanca in order to determine the Percentage of Development of the Sign Language Communication Skill (SLCS) and others within a Collaborative Learning Environment with Mixed-Reality (CLEMR).

Hybridizing metric learning and case-based reasoning for adaptable clickbait detection

The term clickbait is usually used to name web contents which are specifically designed to maximize advertisement monetization, often at the expense of quality and exactitude. The rapid proliferation of this type of content has motivated researchers to develop automatic detection methods, to effectively block clickbaits in different application domains. In this paper, we introduce a novel clickbait detection method. Our approach leverages state-of-the-art techniques from the fields of deep learning and metric learning, integrating them into the Case-Based Reasoning methodology. This provides the model with the ability to learn-over-time, adapting to different users’ criteria. Our experimental results also evidence that the proposed approach outperforms previous clickbait detection methods by a large margin.

Industrial Cyber-Physical Systems in Textile Engineering

Cyber-Physical Systems (CPS) is an emergent approach of physical processes, computer and networking, that focuses on the interaction between cyber and physical elements. These systems monitor and control the physical infrastructures, that is why they have a high impact in industrial automation. The implementation and operation of CPS just like the management of the resulting automation infrastructure is of key importance to the industry. The evolution towards Industry 4.0 is mainly based on digital technologies. We present the integration of Industry 4.0 within the textile industry.

A small look at the ear recognition process using a hybrid approach

The purpose of this document is to offer a combined approach in biometric analysis field, integrating some of the most known techniques using ears to recognize people. This study uses Hausdorff distance as a pre-processing stage adding sturdiness to increase the performance filtering for the subjects to use it in the testing process. Also includes the Image Ray Transform (IRT) and the Haar based classifier for the detection step. Then, the system computes Speeded Up Robust Features (SURF) and Linear Discriminant Analysis (LDA) as an input of two neural networks to recognize a person by the patterns of its ear. To show the applied theory experimental results, the above algorithms have been implemented using Microsoft C#. The investigation results showed robustness improving the ear recognition process.

Data-independent Random Projections from the feature-map of the Homogeneous Polynomial Kernel of degree two

Abstract This paper presents a novel non-linear extension of the Random Projection method based on the degree-2 homogeneous polynomial kernel. Our algorithm is able to implicitly map data points to the high-dimensional feature space of that kernel and from there perform a Random Projection to an Euclidean space of the desired dimensionality. Pairwise distances between data points in the kernel feature space are approximately preserved in the resulting representation. As opposed to previous kernelized Random Projection versions, our method is data-independent and preserves much of the computational simplicity of the original algorithm. This is achieved by focusing on a specific kernel function, what allowed us to analyze the effect of its associated feature mapping in the distribution of the Random Projection hyperplanes. Finally, we present empirical evidence that the proposed method outperforms alternative approaches in terms of pairwise distance preservation, while being significantly more efficient. Also, we show how our method can be used to approximate the accuracy of non-linear classifiers with efficient linear classifiers in some datasets.

A CBR System for Efficient Face Recognition Under Partial Occlusion

This work focuses on the design and validation of a CBR system for efficient face recognition under partial occlusion conditions. The proposed CBR system is based on a classical distance-based classification method, modified to increase its robustness to partial occlusion. This is achieved by using a novel dissimilarity function which discards features coming from occluded facial regions. In addition, we explore the integration of an efficient dimensionality reduction method into the proposed framework to reduce computational cost. We present experimental results showing that the proposed CBR system outperforms classical methods of similar computational requirements in the task of face recognition under partial occlusion.

Minecraft as a Tool in the Teaching-Learning Process of the Fundamental Elements of Circulation in Architecture

In this paper, we make a pedagogical proposal to study the basic elements of circulation in architecture undergraduate degrees, using the Minecraft game as a tool. The theoretical basis for the proposal are Vygotsky’s sociointeractional theory and Ausubel’s theory of meaningful learning. We find ourselves reflecting on Information and Communication Technologies (ICT), gamification and video games in education. We outline some basic elements about circulation in buildings, its types, functionality, and accessibility, and the creativity needed to solve circulation problems in architecture. We introduce the Minecraft game, its characteristics, elements and use it as educational tool. We conclude that video games, specifically Minecraft, are of high interest in education as they develop skills for problem solving, collaborative work, research motivation and proactivity.

Neural Networks Applied to Fingerprint Recognition

In this paper we use a Multi-layer perceptron neural network with learning algorithm retropropagation errors, for application in fingerprint recognition. The objective is to measure the efficiency of the neural network by varying the test data. We observe the behavior of the network in the special case of a partial print. Once the overall structure of the network was designed, tested and properly trained, we proceeded with the testing process, varying the characteristic points and their particular characteristics. Overall, the results demonstrate a stronger recognition when all the characteristic points for the individual prints are available. The recognition rate begins to decrease as the number of characteristic points is reduced to 12, but increases when the number of points is 10, 8 or 5. We obtained a good percentage of hits to remove the features that depended on the center of the footprint and the footprint of the code, in this way to reach the desired goal.