Persona:
Castillo Cara, José Manuel

Foto de perfil
Dirección de correo electrónico
ORCID
0000-0002-2990-7090
Fecha de nacimiento
Proyectos de investigación
Unidades organizativas
Puesto de trabajo
Apellidos
Castillo Cara
Nombre de pila
José Manuel
Nombre

Filtros

2022 - 20244
Restablecer filtros

Ajustes

Autor: García Castro, Raúl × search.filters.applied.f.itemFacultad: E.T.S. de Ingeniería Informática ×

Resultados de la búsqueda

Mostrando 1 - 4 de 4
  • Miniatura
    Publicación
    Generative Adversarial Networks for text-to-face synthesis & generation: A quantitative–qualitative analysis of Natural Language Processing encoders for Spanish
    (Elsevier, 2024-01) Yauri Lozano, Eduardo; Orozco Barbosa, Luis; García Castro, Raúl; Castillo Cara, José Manuel
    In recent years, the development of Natural Language Processing (NLP) text-to-face encoders and Generative Adversarial Networks (GANs) has enabled the synthesis and generation of facial images from textual description. However, most encoders have been developed for the English language. This work presents the first study of three text-to-face encoders, namely, the RoBERTa pre-trained model and the Sent2Vec and RoBERTa models, trained with the CelebA dataset in Spanish. It then introduces customised and fine-tuned conditional Deep Convolutional Generative Adversarial Networks (cDCGANs) trained with the CelebA dataset for text-to-face generation in Spanish. To validate the results obtained, a qualitative evaluation was carried out with a visual analysis and a quantitative evaluation based on the IS, FID and LPIPS metrics. Our findings show promising results with respect to the literature, improving the numerical metrics of FID and LPIPS by 5% and 37%, respectively. Our results also show, through a quantitative–qualitative comparison of the cDCGAN training epochs, that the IS metric is not a reliable objective metric to be considered in the evaluation of similar works
  • Miniatura
    Publicación
    TINTO: Converting Tidy Data into image for classification with 2-Dimensional Convolutional Neural Networks
    (Elsevier, 2023) Talla Chumpitaz, Reewos; García Castro, Raúl; Orozco Barbosa, Luis; Castillo Cara, José Manuel
    The growing interest in the use of algorithms-based machine learning for predictive tasks has generated a large and diverse development of algorithms. However, it is widely known that not all of these algorithms are adapted to efficient solutions in certain tidy data format datasets. For this reason, novel techniques are currently being developed to convert tidy data into images with the aim of using Convolutional Neural Networks (CNNs). TINTO offers the opportunity to convert tidy data into images through the representation of characteristic pixels by implementing two dimensional reduction algorithms: Principal Component Analysis (PCA) and t-distributed Stochastic Neighbour Embedding (t-SNE). Our proposal also includes a blurring technique, which adds more ordered information to the image and can improve the classification task in CNNs.
  • Miniatura
    Publicación
    A novel deep learning approach using blurring image techniques for Bluetooth-based indoor localisation
    (Elsevier, 2022-10-17) Talla Chumpitaz, Reewos; Orozco Barbosa, Luis; García Castro, Raúl; Castillo Cara, José Manuel
    The growing interest in the use of IoT technologies has generated the development of numerous and diverse applications. Many of the services provided by the applications are based on knowledge of the localisation and profile of the end user. Thus, the present work aims to develop a system for indoor localisation prediction using Bluetooth-based fingerprinting using Convolutional Neural Networks (CNN). For this purpose, a novel technique was developed that simulates the diffusion behaviour of the wireless signal by transforming tidy data into images. For this transformation, we implemented the technique used in painting known as blurring, simulating the diffusion of the signal spectrum. Our proposal also includes the use and a comparative analysis of two dimensional reduction algorithms, PCA and t -SNE. Finally, an evolutionary algorithm was implemented to configure and optimise our solution with the combination of different transmission power levels. The results reported in this work present an accuracy of close to 94%, which clearly shows the great potential of this novel technique in the development of more accurate indoor localisation systems .
  • Miniatura
    Publicación
    From cloud and fog computing to federated-fog computing: A comparative analysis of computational resources in real-time IoT applications based on semantic interoperability
    (ELSEVIER, 2024-05-10) Huaranga, Edgar; González Gerpe, Salvador; Castillo Cara, José Manuel; Cimmino, Andrea; García Castro, Raúl; https://orcid.org/0000-0002-8087-0940; https://orcid.org/0000-0003-1550-0430; https://orcid.org/0000-0002-1823-4484; https://orcid.org/0000-0002-0421-452X
    In contemporary computing paradigms, the evolution from cloud computing to fog computing and the recent emergence of federated-fog computing have introduced new challenges pertaining to semantic interoperability, particularly in the context of real-time applications. Fog computing, by shifting computational processes closer to the network edge at the local area network level, aims to mitigate latency and enhance efficiency by minimising data transfers to the cloud. Building upon this, federated-fog computing extends the paradigm by distributing computing resources across diverse organisations and locations, while maintaining centralised management and control. This research article addresses the inherent problematics in achieving semantic interoperability within the evolving architectures of cloud computing, fog computing, and federated-fog computing. Experimental investigations are conducted on a diverse node-based testbed, simulating various end-user devices, to emphasise the critical role of semantic interoperability in facilitating seamless data exchange and integration. Furthermore, the efficacy of federated-fog computing is rigorously evaluated in comparison to traditional fog and cloud computing frameworks. Specifically, the assessment focuses on critical factors such as latency time and computational resource utilisation while processing real-time data streams generated by Internet of Things (IoT) devices. The findings of this study underscore the advantages of federated-fog computing over conventional cloud and fog computing paradigms, particularly in the realm of real-time IoT applications demanding high performance (lowering CPU usage to 20%) and low latency (with picks up to 300ms). The research contributes valuable insights into the optimisation of processing architectures for contemporary computing paradigms, offering implications for the advancement of semantic interoperability in the context of emerging federated-fog computing for IoT applications.