Donostia International Physics Center
Archivo Digital para la Docencia y la InvestigaciónNot a member yet
46642 research outputs found
Sort by
Advancing Face Analysis in Images and Videos: Age Estimation and Drowsiness Detection.
No full text149 p.Deep learning, particularly through sophisticated architectures like CNN, has significantly advanced automated facial analysis. Tasks such as recognition and attribute analysis have seen performance boosts. However, achieving truly robust and versatile systems, especially for complex regression tasks like age estimation or dynamic state assessments like driver drowsiness detection, faces persistent challenges. A key hurdle remains developing models that reliably handle extreme variations in real-world conditions including pose, illumination, expression, occlusions, and intrinsic image quality issues. While techniques like attention mechanisms and specialized network designs exist, accurately interpreting subtle age-related facial changes across a lifetime or detecting fine-grained behavioral cues indicative of drowsiness under these variations remains difficult. Furthermore, ensuring robust generalization across diverse demographics, unseen environments, and varying data acquisition setups often requires more than standard data augmentation or transfer learning, demanding tailored methodological innovations.Addressing these specific challenges in facial age estimation and driver drowsiness detection is crucial. Current age estimation methods, often relying on direct regression or simple classification with standard CNNs, can struggle with the non¬ linear nature of aging, sensitivity to variations unrelated to age, and may not adequately capture distinct features relevant to different life stages. Similarly vision-based drowsiness detection often relies on indicators like PERCLOS, yawn frequency, or head pose, typically extracted using conventional computer vision techniques or basic deep learning models. These approaches can be sensitive to illumination changes, fail to integrate multiple cues effectively, lack robustness to individual differences in fatigue expression, and may not fully leverage the rich spatiotemporal information present in video sequences. Existing methods often lack mechanisms to specifically enhance feature discriminability for these challenging tasks orto compare systematically foundational approaches (handcrafted features) against modern deep learning within these specific contexts.This thesis delves into these specific problems within automated facial analysis, proposing and evaluating advanced computational methods focused explicitly on enhancing facial age estimation from static images and driver drowsiness detection from video sequences. Motivated by the limitations of existing approaches and the need for systems robust to real-world variability, our work explores both the comparative efficacy of traditional handcrafted featu res versus contemporary deep learning techniques and introduces novel deep learning architectures and strategies tailored to these tasks. The primary objective is to develop methods that push the boundaries of accuracy, robustness, and practica! applicability in these domains. The core contributions, validated through extensive experiments detailed herein using relevant benchmark datasets and evaluation metrics (e.g., Mean Absolute Error for age; Accuracy, Fl -seore for drowsiness), are:A comprehensive comparative study systematically evaluating the performance of established handcrafted features against various deep learning-based features for human facial age estimation. This establishes critica! baselines and contextualizes the performance gains achievable with deep learning, while also highlighting scenarios where traditional methods remain competitive or complementary.The development and validation of a novel multi-stage deep neural network architecture specifically designed for facial age estimation. This approach aims to improve accuracy and robustness by decomposing the complex regression task into distinct stages, potentially better modeling age-related transformations across the lifespan.The design and implementation of a specialized Spatiotemporal Convolutional Neural Network (ST-CNN) incorporating Pyramid Bottleneck Blocks. This architecture is demonstrated effectively for eye blinking detection, targeting the efficient capture of multi-scale spatiotemporal features crucial for recognizing micro-expressions relevant to drowsiness analysis in video data.A new hybrid approach for end-to-end driver drowsiness detection in video sequences. This method utilizes a strategy for selecting and integrating deep features from different network levels or temporal windows, aiming to enhance the discriminative power of the feature representation and improve classification performance by focusing on the most salient fatigue indicators overtime.The methodologies employed span comparative analysis, feature engineering, the design of novel deep network architectures (multi-stage CNNs, ST-CNNs with specialized blocks), and hybrid feature selection strategies within deep learning frameworks. Collectively, this research advances the state-of-the-art in robust facial age estimation and video-based drowsiness detection. lt contributes valuable comparative insights and introduces tailored deep learning solutions specifically designed to address the limitations of prior methods and overcome persistent challenges encountered in real-world facial analysis application
Analisis del impacto de la publicidad en redes sociales sobre la percepción corporal y los TCA
No full textEl presente Trabajo de Fin de Grado analiza el impacto de la publicidad en redes sociales sobre la percepción corporal y la aparición de Trastornos de la Conducta Alimentaria (TCA) en jóvenes-adultos. A través de una metodología mixta que combina análisis cuantitativo, entrevistas cualitativas y estudio de campañas publicitarias, se examina cómo los estándares de belleza promovidos digitalmente afectan la autoestima, la autoimagen y los hábitos alimentarios. Se identifican factores de riesgo y se evalúa el papel de los influencers, el uso de filtros digitales y los movimientos de aceptación corporal. El trabajo propone estrategias educativas y regulatorias para disminuir los efectos negativos y fomentar una representación corporal más diversa y realista
Procesos de construcción y negociación de la identidad social en la inmigración marroquí del Pais Vasco
No full text357 p.El objetivo de la presente tesis doctoral, cuyo título es Procesos de construcción y negociación de la identidad social en la inmigración marroquí del País Vasco, consiste en estudiar y profundizar en el conjunto de dinámicas y procesos de construcción y negociación de la identidad social en la comunidad marroquí residente en el País Vasco. Para este fin, se ha optado por un diseño metodológico de carácter cualitativo, siendo la entrevista en profundidad la técnica de investigación empleada. Los principales hallazgos de la investigación se pueden resumir en lo siguiente: la cuestión de la identidad es una realidad social compleja, con articulaciones simbólicas diversas en función de la pertenencia generacional (la primera generación y la generación 1.5) y el género (hombres y mujeres). Al respecto, se ha hallado que las trayectorias sociales implican significados diferentes en torno a la cultura y la religión de origen y de destino. Los miembros de la primera generación, la mayoría hombres, perciben la identidad de una manera más aproximada al origen; mientras que las nuevas generaciones (1.5), más plurales, muestran actitudes de mayor apertura a la sociedad de destino y a la cultura del País Vasco. En concordancia con esto, la actitud hacía la integración social y la identidad se polariza entre una visión instrumental y de recurso y otra de apertura y oportunidad. El riesgo de esta última se encuentra en la tensión que ejerce la no pertenencia exclusiva a una identidad en concreto, lo que puede llevar a una confrontación interna y a un malestar psicológico relevante. Finalmente, los resultados de esta tesis doctoral son relevantes a nivel académico ya que abren un campo de estudio en el País Vasco, además de revelar una realidad social cada vez más importante como es la llegada, incorporación e integración social de las personas inmigrantes marroquíes
Diagnóstico de la infección asociada a ventriculostomía y manejo de la hemorragia intraventricular en el paciente crítico
No full text164 p.Introducción: La infección asociada a ventriculostomía (IAV) es de difícil diagnóstico y no tiene una definición universal. La mortalidad predicha de la hemorragia intraventricular (HIV) alcanza el 80%. Hipótesis: Existen parámetros del líquido cefalorraquídeo (LCR) que diferencian el paciente con IAV. La fibrinolisis intraventricular (FIV) mejora el pronóstico de la HIV sin aumentar significativamente la IAV.Objetivos: 1 Evaluar la utilidad de la citobioquímica (CTBQ) del LCR para diagnosticar la IAV. 2 Describir los factores de riesgo de la IAV. 3 Analizar el pronóstico del paciente con HIV tratado con FIV Metodología: análisis retrospectivo de 18 años con todos los pacientes con drenaje ventricular externo (DVE). Se compararon las variables de pacientes infectados frente a no infectados. Se calcularon la sensibilidad, especificidad y valores predictivos (VP) para cada parámetro diagnóstico. Se realizó unanálisis de regresión logística multivariante para detectar los factores de riesgo y uno de regresión logística multinomial para predecir el resultado del cultivo del LCR. Nivel de significación 42/µL, proteinorraquia > a 118,5 mg/dL e índice celular >15850/µL, mostraron un VP negativo (VPN) >95%. Se construyó un modelo predictivo del resultado del cultivo de LCR con variables clínicas, los cultivos previos y la CTBQ del LCR. En los pacientes con HIV moderada/grave la incidencia fue 22,8% (15,2 IAV por mil días de DVE). Fallecieron el 40,2%. El APACHE y la antiagregación o anticoagulación se asociaron con la mortalidad.Conclusiones: la CTBQ del LCR tiene alto VPN diagnóstico de la IAV. Un modelo predictor del resultado del cultivo puede anticipar este diagnóstico. La HIV tratada con FIV tiene < mortalidad que la predich
Enhancement of melanoma aggressiveness via p38-MAPK, HIF-1α pathways, and metabolic reprogramming induced by Candida albicans
No full textRecent studies have increasingly focused on the role of fungi, including Candida albicans, in carcinogenesis. Since C. albicans is a component of the human microbiota, particularly on the skin, we investigated its effect on the phenotype and signalling pathways of melanoma cells. Assays for migration, adhesion, angiogenesis, and hepatic metastasis showed that C. albicans promotes a more malignant phenotype in melanoma cells. At the transcriptomic level, C. albicans increased the expression of VEGF (Vegfa), and genes associated with MAPK and HIF-1 signalling pathways, and with aerobic glycolysis. Further in vitro analysis revealed that TLRs and EphA2 receptors are involved in the recognition of live C. albicans, stimulating VEGF secretion and expression of the AP-1 transcription factor component c-Fos through p38-MAPK and HIF-1α. These pathways also regulate the expression of other AP-1 constituents such as Atf3, Jun, and Jund. Moreover, p38-MAPK regulates glycolytic genes like Hk2, Slc2a1, and Eno2. In conclusion, C. albicans activates the p38-MAPK/c-Fos/AP-1 and HIF-1/HIF-1α/c-Fos/AP-1 pathways in melanoma cells, promoting a pro-angiogenic environment and metabolic reprogramming. Therefore, this study clarifies the impact of C. albicans on melanoma cells, which can lead to the use of antifungal therapies as complementary to traditional treatments for melanoma.Basque Government (Grant numbers IT1362-19 and IT1657-22).
Basque Department of Industry, Tourism and Trade (Etortek, Elkartek and Emaitek Programs),
The Innovation Technology Department of Bizkaia County, CIBERehd Network, and Spanish MINECO the Severo Ochoa Excellence Accreditation [CEX2021-001136-S]
Use of context in updating affective representations of words in older adults
No full textAvailable online 22nd February, 2025.Older adults (OAs) often prioritize positive over negative information during word processing, termed as positivity bias. However, it is unclear how OAs update the affective representation of a word in contexts. The present study examined whether age-related positivity bias influences the update of the affective representation of a word in different emotional contexts. In Experiment 1 (web-based), younger and older participants read positive and negative target words in positive and negative contexts and rated the valence of the target words. Negative contexts biased the ratings more than positive ones, reflecting a negativity bias during offline valence evaluation in both age groups. In Experiment 2 (EEG), another group of participants read positive and negative target words in positive and negative contexts first, and then the same target words again, and made valence judgment on the target words. OAs showed a larger P2 (180–300 ms) difference before and after contexts for positive target words than younger adults (YAs). This suggests OAs’ early attention to positive features of words in contexts. YAs showed a larger late positive complex (LPC) difference for target words before and after negative contexts than before and after positive contexts, while older adults showed comparable LPC effects across all the conditions. This suggests that YAs use negative contexts to update the affective representation of a word, whereas OAs do so in both positive and negative contexts. Our findings supported a reduced negativity bias in OAs in using (emotional) contexts to update the affective neural representation of a word.This work was supported by the National Science Foundation (Grant no. BCS 2213622, EES 1619524), William Orr Dingwall Dissertation Fellowship in the Cognitive, Clinical and Neural Foundations of Language, and 2022 National Science and Technology Council Taiwanese Overseas Pioneers Grants for PhD Candidates
Charting Global Language Regulation Practices towards Language Maintenance and Revitalisation
No full text302 p.This thesis investigates the work of language regulators dedicated to endangered languages, focusing on their maintenance, promotion and revitalisation strategies, as well as their organizational structures. It addresses the gap in the literature concerning the practical operations of such institutions. Conducted on a global scale, the dissertation aims to develop a typology of regulatory practices, spanning from macro-level strategies like formal education to micro-level ones such as dictionary creation and digitalisation. The methodology concerns a questionnaire sent to a wide range of language regulators, collecting data on their roles, activities, status and general revitalisation strategies. The languages that qualify as endangered were selected via the UNESCO¿s World Atlas of Endangered Languages and the regulators were identified through targeted online search. The collected data was organised in a spreadsheet format for comparative analysis between various variables. This method identified patterns and typologies in regulatory approaches. Finally, the dissertation contributes significantly to the field by creating an unprecedented global resource on endangered language regulators and by proposing actionable strategies to benefit such linguistic communities and for effective language preservatio
Designing learning tools in the AI Era: Developing for design science research in PhD education
No full text194 p.Traditional learning theories often view learning as an integrative process where learners must connect new information to their existing knowledge. The increasing presence of Artificial Intelligence (AI) technologies, particularly Large Language Models (LLMs), is transforming the landscape of learning tools by shifting from traditional reading towards interactive prompting techniques. From a learning perspective, doctoral studies differs significantly from previous education due to independent research, creation of new knowledge, and advanced critical thinking, with minimal direct instruction. Moreover, when resorting to Design Science Research (DSR) as methodology, students typically have difficulties in capturing and representing the complexities of practical problems and understanding the opportunities and constraints necessary for impactful contributions to practice.Using Action Design Research (ADR) and guided by a 2x2 classification matrix, which organizes knowledge representation into new information (documents or LLMs) and existing knowledge (text or diagrams), this thesis focuses on supporting three key activities in doctoral studies: knowledge representation, problem scoping, and writing. This thesis resulted in the design of three web extensions: Concept&Go, Chatin, and PrompTeX.This work was hosted by the University of the Basque Country (Faculty of Informatics). The author enjoyed
a doctoral grant from the University of the Basque Country under the contract - PIF20/236 from 2020
to 2025. The work was co-supported by MCIN/AEI/10.13039/501100011033/FEDER, UE, the “European
Union NextGenerationEU/PRTR” under contract PID2021-125438OB-I00 and the program “Convocatoria
para la Concesion de Ayudas a los Grupos de Investigacion en la Universidad del Pais Vasco/Euskal Herriko Unibertsitatea (2021)” under the contract GIU21/03
Transformation and continuity in the urban regimes of Bilbao and Barcelona: A comparative analysis using urban regime theory
No full textThis article examines transformation and continuity in the urban regimes of Bilbao and Barcelona from the outbreak of the 15-M movement to the present day, using the urban regime theory as an analytical framework. This article argues that despite the difficulties in establishing a sustained and stable urban regime over time, the municipalist cycle in Barcelona has produced an alternative regulative idea, signaling an attempt at regime change that, while not a total rupture, lays the foundations for a new urban regime. In contrast, we present the case of Bilbao where, through small arrangements, the same urban regime based on renewed business strategies is sustained and reproduced thanks to the hegemony of the ruling party. In this way, we explore the power dynamics and adaptive responses of urban regimes to the waves of neoliberal destruction and creation. By studying the different structures and strategic selectivities, the article puts forward new relevant analytical tools for comparing similar cases in different contexts. Thus, through a theoretical review, the study provides a deeper understanding of the challenges faced by urban regimes in post-industrial cities like Barcelona and Bilbao
