DI - CAOS - Artículos de Revistas

Permanent URI for this collection


Recent Submissions

Now showing 1 - 20 of 44
  • Publication
    Stress Detection Using Frequency Spectrum Analysis of Wrist-Measured Electrodermal Activity
    (MDPI, 2023-01-01) Stržinar, Žiga; Sanchis de Miguel, María Araceli; Ledezma Espino, Agapito Ismael; Sipele, Oscar; Pregelj, Boštjan; Skrjanc, Igor; Banco Santander; Ministerio de Economía y Competitividad (España); Universidad Carlos III de Madrid
    The article deals with the detection of stress using the electrodermal activity (EDA) signal measured at the wrist. We present an approach for feature extraction from EDA. The approach uses frequency spectrum analysis in multiple frequency bands. We evaluate the proposed approach using the 4 Hz EDA signal measured at the wrist in the publicly available Wearable Stress and Affect Detection (WESAD) dataset. Seven existing approaches to stress detection using EDA signals measured by wrist-worn sensors are analysed and the reported results are compared with ours. The proposed approach represents an improvement in accuracy over the other techniques studied. Moreover, we focus on time to detection (TTD) and show that our approach is able to outperform competing techniques, with fewer data points. The proposed feature extraction is computationally inexpensive, thus the presented approach is suitable for use in real-world wearable applications where both short response times and high detection performance are important. We report both binary (stress vs. no stress) as well as three-class (baseline/stress/amusement) results.
  • Publication
    Iot application for energy poverty detection based on thermal comfort monitoring
    (Elsevier, 2023-01-14) López Vargas, Ascensión; Ledezma Espino, Agapito Ismael; European Commission; Ministerio de Ciencia e Innovación (España)
    The development of a datalogger for identifying Energy Poverty (EP) using thermal comfort monitoring is described in this work. There is not a uniform definition of EP, and no global recommendations indicating the thermal comfort characteristics that should be utilized to identify EP. Most Internet of Things (IoT)-based systems designed for EP identification measure energy consumptions (electricity and gas). There is a lack of works that use IoT-based systems to identify EP through the monitoring of thermal comfort parameters. To address the deficiencies discovered in the identification of EP from the perspective of thermal efficiency, an IoT-based monitoring system was designed, developed, and tested. A first pilot was installed in a household in Getafe. A full month of temperature, relative humidity, and CO2 concentration measurements were utilized to evaluate the system, which was then compared to a commercial system. The results revealed that the new IoT-based approach was very dependable and may be used to accurately monitor EP-related parameters.
  • Publication
    Automated driving: A literature review of the take over request in conditional automation
    (MDPI, 2020-12) Morales-Álvarez, Walter; Sipele, Oscar; Leberon, Regis; Tadjine, Hadj Hamma; Olaverri-Monreal, Cristina; Ministerio de Economía y Competitividad (España)
    In conditional automation (level 3), human drivers can hand over the Driving Dynamic Task (DDT) to the Automated Driving System (ADS) and only be ready to resume control in emergency situations, allowing them to be engaged in non-driving related tasks (NDRT) whilst the vehicle operates within its Operational Design Domain (ODD). Outside the ODD, a safe transition process from the ADS engaged mode to manual driving should be initiated by the system through the issue of an appropriate Take Over Request (TOR). In this case, the driver's state plays a fundamental role, as a low attention level might increase driver reaction time to take over control of the vehicle. This paper summarizes and analyzes previously published works in the field of conditional automation and the TOR process. It introduces the topic in the appropriate context describing as well a variety of concerns that are associated with the TOR. It also provides theoretical foundations on implemented designs, and report on concrete examples that are targeted towards designers and the general public. Moreover, it compiles guidelines and standards related to automation in driving and highlights the research gaps that need to be addressed in future research, discussing also approaches and limitations and providing conclusions.
  • Publication
    Explaining Deep Learning-Based Driver Models
    (MDPI, 2021-04-07) Sesmero Lorente, María Paz; Magan Lopez, Elena; Álvarez Flórez, Laura; Ledezma Espino, Agapito Ismael; Iglesias Martínez, José Antonio; Sanchis de Miguel, María Araceli; Comunidad de Madrid
    Different systems based on Artificial Intelligence (AI) techniques are currently used in relevant areas such as healthcare, cybersecurity, natural language processing, and self-driving cars. However, many of these systems are developed with 'black box” AI, which makes it difficult to explain how they work. For this reason, explainability and interpretability are key factors that need to be taken into consideration in the development of AI systems in critical areas. In addition, different contexts produce different explainability needs which must be met. Against this background, Explainable Artificial Intelligence (XAI) appears to be able to address and solve this situation. In the field of automated driving, XAI is particularly needed because the level of automation is constantly increasing according to the development of AI techniques. For this reason, the field of XAI in the context of automated driving is of particular interest. In this paper, we propose the use of an explainable intelligence technique in the understanding of some of the tasks involved in the development of advanced driver-assistance systems (ADAS). Since ADAS assist drivers in driving functions, it is essential to know the reason for the decisions taken. In addition, trusted AI is the cornerstone of the confidence needed in this research area. Thus, due to the complexity and the different variables that are part of the decision-making process, this paper focuses on two specific tasks in this area: the detection of emotions and the distractions of drivers. The results obtained are promising and show the capacity of the explainable artificial techniques in the different tasks of the proposed environments.
  • Publication
    IoT for Global Development to Achieve the United Nations Sustainable Development Goals: The New Scenario After the COVID-19 Pandemic
    (IEEE, 2021-08-01) López Vargas, Ascención; Ledezma Espino, Agapito Ismael; Bott, Jack; Sanchis de Miguel, María Araceli; European Commission; Ministerio de Ciencia e Innovación (España)
    COVID-19 has not affected all countries equally: developing countries have been more disadvantaged by the pandemic. Regarding global development, the COVID-19 pandemic has forced a step back in the path to attaining the Sustainable Development Goals (SDGs). The SDGs most negatively affected by the pandemic are identified here: education, health, and work. Then using the SDGs as a reference, this research explores the new challenges faced by developing countries and the impact of the Internet of Things (IoT) after COVID-19's emergence. IoT solutions carried out in developing countries during the pandemic have been identified and reviewed. Successful Internet of Things for Development (IoT4D) projects, in relation to the SDGs, are highlighted. New social and technical challenges that have emerged for the IoT4D as a consequence of the pandemic are then studied. This work concludes that the future of IoT4D in the wake of COVID-19 should focus on the use of low-cost IoT devices for the SDGs most affected by the pandemic. After an exhaustive study, the Intelligent Internet of Things (IIoT) has been determined to be a key actor in the pandemic's wake, with a leading role in the health sector. The proposed approach includes an extensive study of the new role of the IoT4D for achieving the SDGs in our forever changed world.
  • Publication
    Detection of kinase domain mutations in BCR::ABL1 leukemia by ultra-deep sequencing of genomic DNA
    (Nature Research, 2022-07-29) Sanchez, Ricardo; Dorado Alfaro, Sara; Toledo Heras, María Paula de
    The screening of the BCR::ABL1 kinase domain (KD) mutation has become a routine analysis in case of warning/failure for chronic myeloid leukemia (CML) and B-cell precursor acute lymphoblastic leukemia (ALL) Philadelphia (Ph)-positive patients. In this study, we present a novel DNA-based next-generation sequencing (NGS) methodology for KD ABL1 mutation detection and monitoring with a 1.0E−4 sensitivity. This approach was validated with a well-stablished RNA-based nested NGS method. The correlation of both techniques for the quantification of ABL1 mutations was high (Pearson r = 0.858, p < 0.001), offering DNA-DeepNGS a sensitivity of 92% and specificity of 82%. The clinical impact was studied in a cohort of 129 patients (n = 67 for CML and n = 62 for B-ALL patients). A total of 162 samples (n = 86 CML and n = 76 B-ALL) were studied. Of them, 27 out of 86 harbored mutations (6 in warning and 21 in failure) for CML, and 13 out of 76 (2 diagnostic and 11 relapse samples) did in B-ALL patients. In addition, in four cases were detected mutation despite BCR::ABL1 < 1%. In conclusion, we were able to detect KD ABL1 mutations with a 1.0E−4 sensitivity by NGS using DNA as starting material even in patients with low levels of disease.
  • Publication
    Evolving Gaussian on-line clustering in social network analysis
    (Elsevier, 2022-11-30) Skrjanc, Igor; Andonovski, Goran; Iglesias Martínez, José Antonio; Sesmero Lorente, María Paz; Sanchis de Miguel, María Araceli; Comunidad de Madrid; Ministerio de Economía y Competitividad (España)
    In this paper, we present an evolving data-based approach to automatically cluster Twitter users according to their behavior. The clustering method is based on the Gaussian probability density distribution combined with a Takagi-Sugeno fuzzy consequent part of order zero (eGauss0). This means that this method can be used as a classifier that is actually a mapping from the feature space to the class label space. The eGauss method is very flexible, is computed recursively, and the most important thing is that it starts learning 'from scratch'. The structure adapts to the new data using adding and merging mechanisms. The most important feature of the evolving method is that it can process data from thousands of Twitter profiles in real time, which can be characterized as a Big Data problem. The final clusters yield classes of Twitter profiles, which are represented as different activity levels of each profile. In this way, we could classify each member as ordinary, very active, influential and unusual user. The proposed method was also tested on the Iris and Breast Cancer Wisconsin datasets and compared with other methods. In both cases, the proposed method achieves high classification rates and shows competitive results.
  • Publication
    MCO plan: Efficient coverage mission for multiple micro aerial vehicles modeled as agents
    (MDPI, 2022-07) Campo, Liseth Viviana; Ledezma Espino, Agapito Ismael; Corrales, Juan Carlos
    Micro aerial vehicle (MAV) fleets have gained essential recognition in the decision schemes for precision agriculture, disaster management, and other coverage missions. However, they have some challenges in becoming massively deployed. One of them is resource management in restricted workspaces. This paper proposes a plan to balance resources when considering the practical use of MAVs and workspace in daily chores. The coverage mission plan is based on five stages: world abstraction, area partitioning, role allocation, task generation, and task allocation. The tasks are allocated according to agent roles, Master, Coordinator, or Operator (MCO), which describe their flight autonomy, connectivity, and decision skill. These roles are engaged with the partitioning based on the Voronoi-tessellation but extended to heterogeneous polygons. The advantages of the MCO Plan were evident compared with conventional Boustrophedon decomposition and clustering by K-means. The MCO plan achieved a balanced magnitude and trend of heterogeneity between both methods, involving MAVs with few or intermediate resources. The resulting efficiency was tested in the GAMA platform, with gained energy between 2% and 10% in the mission end. In addition, the MCO plan improved mission times while the connectivity was effectively held, even more, if the Firefly algorithm generated coverage paths.
  • Publication
    BirdNet+: two-stage 3D object detection in LiDAR through a sparsity-invariant bird's eye view
    (IEEE, 2021-11-30) Barrera Del Pozo, Alejandro; Beltrán de la Cita, Jorge; Guindel Gómez, Carlos; Iglesias Martínez, José Antonio; García Fernández, Fernando; Comunidad de Madrid
    Autonomous navigation relies upon an accurate understanding of the elements in the surroundings. Among the different on-board perception tasks, 3D object detection allows the identification of dynamic objects that cannot be registered by maps, being key for safe navigation. Thus, it often requires the use of LiDAR data, which is able to faithfully represent the scene geometry. However, although raw laser point clouds contain rich features to perform object detection, more compact representations such as the bird's eye view (BEV) projection are usually preferred in order to meet the time requirements of the control loop. This paper presents an end-to-end object detection network based on the well-known Faster R-CNN architecture that uses BEV images as input to produce the final 3D boxes. Our regression branches can infer not only the axis-aligned bounding boxes but also the rotation angle, height, and elevation of the objects in the scene. The proposed network provides state-of-the-art results for car, pedestrian, and cyclist detection with a single forward pass when evaluated on the KITTI 3D Object Detection Benchmark, with an accuracy that exceeds 64% mAP 3D for the Moderate difficulty. Further experiments on the challenging nuScenes dataset show the generalizability of both the method and the proposed BEV representation against different LiDAR devices and across a wider set of object categories by being able to reach more than 30% mAP with a single LiDAR sweep and almost 40% mAP with the usual 10-sweep accumulation.
  • Publication
    A case-based reasoning system for recommendation of data cleaning algorithms in classification and regression tasks
    (Elsevier, 2020-02-18) Corrales Muñoz, David Camilo; Ledezma Espino, Agapito Ismael; Corrales, Juan Carlos; Ministerio de Economía, Industria y Competitividad (España)
    Recently, advances in Information Technologies (social networks, mobile applications, Internet of Things, etc.) generate a deluge of digital data; but to convert these data into useful information for business decisions is a growing challenge. Exploiting the massive amount of data through knowledge discovery (KD) process includes identifying valid, novel, potentially useful and understandable patterns from a huge volume of data. However, to prepare the data is a non-trivial refinement task that requires technical expertise in methods and algorithms for data cleaning. Consequently, the use of a suitable data analysis technique is a headache for inexpert users. To address these problems, we propose a case-based reasoning system (CBR) to recommend data cleaning algorithms for classification and regression tasks. In our approach, we represent the problem space by the meta-features of the dataset, its attributes, and the target variable. The solution space contains the algorithms of data cleaning used for each dataset. We represent the cases through a Data Cleaning Ontology. The case retrieval mechanism is composed of a filter and similarity phases. In the first phase, we defined two filter approaches based on clustering and quartile analysis. These filters retrieve a reduced number of relevant cases. The second phase computes a ranking of the retrieved cases by filter approaches, and it scores a similarity between a new case and the retrieved cases. The retrieval mechanism proposed was evaluated through a set of judges. The panel of judges scores the similarity between a query case against all cases of the case-base (ground truth). The results of the retrieval mechanism reach an average precision on judges ranking of 94.5% in top 3, for top 7 84.55%, while in top 10 78.35%.
  • Publication
    Web news mining in an evolving framework
    (Elsevier, 2016-03-01) Iglesias Martínez, José Antonio; Tiemblo, Alexandra; Ledezma Espino, Agapito Ismael; Sanchis de Miguel, María Araceli
    Online news has become one of the major channels for Internet users to get news. News websites are daily overwhelmed with plenty of news articles. Huge amounts of online news articles are generated and updated everyday, and the processing and analysis of this large corpus of data is an important challenge. This challenge needs to be tackled by using big data techniques which process large volume of data within limited run times. Also, since we are heading into a social-media data explosion, techniques such as text mining or social network analysis need to be seriously taken into consideration. In this work we focus on one of the most common daily activities: web news reading. News websites produce thousands of articles covering a wide spectrum of topics or categories which can be considered as a big data problem. In order to extract useful information, these news articles need to be processed by using big data techniques. In this context, we present an approach for classifying huge amounts of different news articles into various categories (topic areas) based on the text content of the articles. Since these categories are constantly updated with new articles, our approach is based on Evolving Fuzzy Systems (EFS). The EFS can update in real time the model that describes a category according to the changes in the content of the corresponding articles. The novelty of the proposed system relies in the treatment of the web news articles to be used by these systems and the implementation and adjustment of them for this task. Our proposal not only classifies news articles, but it also creates human interpretable models of the different categories. This approach has been successfully tested using real on-line news. (C) 2015 Elsevier B.V. All rights reserved.
  • Publication
    CCE: An ensemble architecture based on coupled ANN for solving multiclass problems
    (Elsevier, 2020-06-01) Sesmero Lorente, María Paz; Alonso Weber, Juan Manuel; Sanchis de Miguel, María Araceli; Ministerio de Economía y Competitividad (España)
    The resolution of multiclass classification problems has been usually addressed by using a "divide and conquer" strategy that splits the original problem into several binary subproblems. This approach is mandatory when the learning algorithm has been designed to solve binary problems and a multiclass version cannot be devised. Artificial Neural Networks, ANN, are binary learning models whose extension to multiclass problems is rather straightforward by using the standard 1-out-of N codification of the classes. However, the use of a single ANN can be inefficient in terms of accuracy and computational complexity when the data set is large, or the number of classes is high. In this work, we exhaustively describe CCE, a new classifier ensemble based on ANN. Each member of this new ensemble is a couple of multiclass ANN's. Each ANN is trained using different subsets of the dataset ensuring these subsets to be disjoint. This new approach allows to combine the benefits of the divide and conquer methodology, with the use of multiclass ANNs and with the combination of individual classification modules that give a complete answer to the addressed problem. The combination of these elements results in a classifier ensemble in which the diversity of the base classifiers provides high accuracy values. Moreover, the use of couples of ANN proves to be tolerant to labeling noise and computationally efficient. The performance of CCE has been tested on various datasets and the results show the higher performance of this approach with respect to other used classification systems.
  • Publication
    Autonomously evolving classifier TEDAClass
    (Elsevier, 2016-10-20) Kangin, Dmitry; Angelov, Plamen P.; Iglesias Martínez, José Antonio
    In this paper we introduce a classifier named TEDAClass (Typicality and Eccentricity based Data Analytics Classifier) which is based on the recently proposed AnYa type fuzzy rule based system. Specifically, the rules of the proposed classifier are defined according to the recently proposed TEDA framework. This novel and efficient systematic methodology for data analysis is a promising addition to the traditional probability as well as to the fuzzy logic. It is centred at non-parametric density estimation derived from the data sample. In addition, the proposed framework is computationally cheap and provides fast and exact per-point processing of the data set/stream. The algorithm is demonstrated to be suitable for different classification tasks. Throughout the paper we give evidence of its applicability to a wide range of practical problems. Furthermore, the algorithm can be easily adapted to different classical data analytics problems, such as clustering, regression, prediction, and outlier detection. Finally, it is very important to remark that the proposed algorithm can work "from scratch" and evolve its structure during the learning process. (C) 2016 Elsevier Inc. All rights reserved.
  • Publication
    Fault detection and identification methodology under an incremental learning framework applied to industrial machinery
    (IEEE, 2018-09-04) Carino, Jesús A.; Delgado Prieto, Miguel; Zurita, Daniel; Ortega Redondo, Juan Antonio; Iglesias Martínez, José Antonio; Sanchis de Miguel, María Araceli; Millan, Marta; Romero Troncoso, Rene; Ministerio de Economía y Competitividad (España)
    An industrial machinery condition monitoring methodology based on ensemble novelty detection and evolving classification is proposed in this study. The methodology contributes to solve current challenges dealing with classical electromechanical system monitoring approaches applied in industrial frameworks, that is, the presence of unknown events, the limitation to the nominal healthy condition as starting knowledge, and the incorporation of new patterns to the available knowledge. The proposed methodology is divided into four main stages: 1) a dedicated feature calculation and reduction over available physical magnitudes to increase novelty detection and fault classification capabilities; 2) a novelty detection based on the ensemble of one-class support vector machines to identify not previously considered events; 3) a diagnosis by means of eClass evolving classifiers for patterns recognition; and 4) re-training to include new patterns to the novelty detection and fault identification models. The effectiveness of the proposed fault detection and identification methodology has been compared with classical approaches, and verified by experimental results obtained from an automotive end-of-line test machine.
  • Publication
    Evolving fuzzy and neuro-fuzzy approaches in clustering, regression, identification, and classification: A Survey
    (Elsevier, 2019-07-01) Skrjanc, Igor; Iglesias Martínez, José Antonio; Sanchis de Miguel, María Araceli; Leite, Daniel; Lughofer, Edwin; Gomide, Fernando; Banco Santander; Universidad Carlos III de Madrid
    Major assumptions in computational intelligence and machine learning consist of the availability of a historical dataset for model development, and that the resulting model will, to some extent, handle similar instances during its online operation. However, in many real world applications, these assumptions may not hold as the amount of previously available data may be insufficient to represent the underlying system, and the environment and the system may change over time. As the amount of data increases, it is no longer feasible to process data efficiently using iterative algorithms, which typically require multiple passes over the same portions of data. Evolving modeling from data streams has emerged as a framework to address these issues properly by self-adaptation, single-pass learning steps and evolution as well as contraction of model components on demand and on the fly. This survey focuses on evolving fuzzy rule-based models and neuro-fuzzy networks for clustering, classification and regression and system identification in online, real-time environments where learning and model development should be performed incrementally. (C) 2019 Published by Elsevier Inc.
  • Publication
    Evolving cloud-based system for the recognition of drivers' actions
    (Elsevier, 2018-06-01) Skrjanc, Igor; Andonovski, Goran; Ledezma Espino, Agapito Ismael; Sipele Siale, Botyuu Óscar; Iglesias Martínez, José Antonio; Sanchis de Miguel, María Araceli; Ministerio de Economía y Competitividad (España); Universidad Carlos III de Madrid
    This paper presents an evolving cloud-based algorithm for the recognition of drivers' actions. The general idea is to detect different manoeuvres by processing the standard signals that are usually measured in a car, such as the speed, the revolutions, the angle of the steering wheel, the position of the pedals, and others, without additional intelligent sensors. The primary goal of this investigation is to propose a concept that can be used to recognise various driver actions. All experiments are performed on a realistic car simulator. The data acquired from the simulator are pre-processed and then used in the evolving cloud-based algorithm to detect the basic elementary actions, which are then combined in a prescribed sequence to create tasks. Finally, the sequences of different tasks form the most complex action, which is called a manoeuvre. As shown in this paper, the evolving cloud-based algorithm can be very efficiently used to recognise the complex driver's action from raw signals obtained by typical car sensors. (C) 2017 Elsevier Ltd. All rights reserved.
  • Publication
    Driver Drowsiness Detection by Applying Deep Learning Techniques to Sequences of Images
    (MDPI AG, 2022-02) Magan Lopez, Elena; Sesmero Lorente, María Paz; Alonso Weber, Juan Manuel; Sanchis de Miguel, María Araceli; Comunidad de Madrid; Ministerio de Economía y Competitividad (España); Ministerio de Ciencia, Innovación y Universidades (España)
    This work presents the development of an ADAS (advanced driving assistance system) focused on driver drowsiness detection, whose objective is to alert drivers of their drowsy state to avoid road traffic accidents. In a driving environment, it is necessary that fatigue detection is performed in a non-intrusive way, and that the driver is not bothered with alarms when he or she is not drowsy. Our approach to this open problem uses sequences of images that are 60 s long and are recorded in such a way that the subject’s face is visible. To detect whether the driver shows symptoms of drowsiness or not, two alternative solutions are developed, focusing on the minimization of false positives. The first alternative uses a recurrent and convolutional neural network, while the second one uses deep learning techniques to extract numeric features from images, which are introduced into a fuzzy logic-based system afterwards. The accuracy obtained by both systems is similar: around 65% accuracy over training data, and 60% accuracy on test data. However, the fuzzy logic-based system stands out because it avoids raising false alarms and reaches a specificity (proportion of videos in which the driver is not drowsy that are correctly classified) of 93%. Although the obtained results do not achieve very satisfactory rates, the proposals presented in this work are promising and can be considered a solid baseline for future works.
  • Publication
    Implementing a Gaze Tracking Algorithm for Improving Advanced Driver Assistance Systems
    (MDPI, 2021-06-19) Ledezma Espino, Agapito Ismael; Zamora España, Víctor Manuel; SIpele Siale; Sesmero Lorente, María Paz; Sanchis de Miguel, María Araceli; Comunidad de Madrid; Ministerio de Ciencia e Innovación (España)
    Car accidents are one of the top ten causes of death and are produced mainly by driver distractions. ADAS (Advanced Driver Assistance Systems) can warn the driver of dangerous scenarios, improving road safety, and reducing the number of traffic accidents. However, having a system that is continuously sounding alarms can be overwhelming or confusing or both, and can be counterproductive. Using the driver"s attention to build an efficient ADAS is the main contribution of this work. To obtain this 'attention value” the use of a Gaze tracking is proposed. Driver"s gaze direction is a crucial factor in understanding fatal distractions, as well as discerning when it is necessary to warn the driver about risks on the road. In this paper, a real-time gaze tracking system is proposed as part of the development of an ADAS that obtains and communicates the driver"s gaze information. The developed ADAS uses gaze information to determine if the drivers are looking to the road with their full attention. This work gives a step ahead in the ADAS based on the driver, building an ADAS that warns the driver only in case of distraction. The gaze tracking system was implemented as a model-based system using a Kinect v2.0 sensor and was adjusted on a set-up environment and tested on a suitable-features driving simulation environment. The average obtained results are promising, having hit ratios between 96.37% and 81.84%
  • Publication
    Lane following learning based on semantic segmentation with chroma key and image superposition
    (MDPI AG, 2021-12) Corrochano Jiménez, Javier; Alonso Weber, Juan Manuel; Sesmero Lorente, María Paz; Sanchis de Miguel, María Araceli; Comunidad de Madrid; Ministerio de Economía y Competitividad (España); Ministerio de Ciencia e Innovación (España)
    There are various techniques to approach learning in autonomous driving; however, all of them suffer from some problems. In the case of imitation learning based on artificial neural networks, the system must learn to correctly identify the elements of the environment. In some cases, it takes a lot of effort to tag the images with the proper semantics. This is also relevant given the need to have very varied scenarios to train and to thus obtain an acceptable generalization capacity. In the present work, we propose a technique for automated semantic labeling. It is based on various learning phases using image superposition combining both scenarios with chromas and real indoor scenarios. This allows the generation of augmented datasets that facilitate the learning process. Further improvements by applying noise techniques are also studied. To carry out the validation, a small-scale car model is used that learns to automatically drive on a reduced circuit. A comparison with models that do not rely on semantic segmentation is also performed. The main contribution of our proposal is the possibility of generating datasets for real indoor scenarios with automatic semantic segmentation, without the need for endless human labeling tasks.
  • Publication
    Impact of the learners diversity and combination method on the generation of heterogeneous classifier ensembles
    (Elsevier, 2021-07-15) Sesmero Lorente, María Paz; Iglesias Martínez, José Antonio; Magan Lopez, Elena; Ledezma Espino, Agapito Ismael; Sanchis de Miguel, María Araceli; Agencia Estatal de Investigación (España)
    Ensembles of classifiers is a proven approach in machine learning with a wide variety of research works. The main issue in ensembles of classifiers is not only the selection of the base classifiers, but also the combination of their outputs. According to the literature, it has been established that much is to be gained from combining classifiers if those classifiers are accurate and diverse. However, it is still an open issue how to define the relation between accuracy and diversity in order to define the best possible ensemble of classifiers. In this paper, we propose a novel approach to evaluate the impact of the diversity of the learners on the generation of heterogeneous ensembles. We present an exhaustive study of this approach using 27 different multiclass datasets and analysing their results in detail. In addition, to determine the performance of the different results, the presence of labelling noise is also considered.