21793 research outputs found
Sort by
Defending Object Detection Models against Image Distortions
Get PDFObject detection has significantly advanced with deep learning but faces challenges under image distortions like noise, compression, blur, fog, and snow. This issue is critical in applications such as self-driving cars and healthcare. While defence methods aim to enhance robustness under distortions, maintaining performance on clean images remains a challenge.
To address this challenge, we propose a novel defence approach that generates copies of the original training images and adds distortion-like content to these copies at the pixel level. We balance the number of distorted pixels to prevent bias during learning. Our approach includes two augmentation methods to generate augmented samples: AISbod, which uses artificial immune systems (AIS), and GSES, which employs kernel density estimation (KDE).
Our AISbod uses AIS to distort the original sample (antigen) through cycles of “select, clone, mutate, select” until the augmented data (antibody) reaches a specified similarity to the antigen. However, AIS is limited in diversifying generated antibodies and is computationally expensive. Therefore, in GSES, we create samples by selecting pixels from the original samples, estimating the pixel distribution from multiple distorted versions of the original samples via KDE, and then replacing the selected pixels with new values sampled from the estimated distribution. GSES generates more diverse data than AISbod.
We evaluate our methods on 15 image distortions using state-of-the-art object detection models like DINO and YOLOv7. Our methods improve accuracy under distorted and clean images and remain consistent across datasets and detection models. For instance, DINO on the COCO dataset shows a 4.50% improvement under clean samples, 8.40% on average across all distortions, 2.50% under snow, and 29.30% under impulse noise. The observed improvement is due to the weight regularization effect of our methods, which is evident in the smoother convergence of training and validation loss curves, indicating reduced learning fluctuations and a more stable optimization path. Additionally, the narrower gap between the curves suggests reduced over-fitting, leading to better generalization to unseen data. Simulations indicate that our approach outperforms related defence methods against distortions and extends beyond object detection, improving accuracy in image classification and object tracking models
From A Rhetoric of Ravishment to A Rhetoric of Rape: Sir Thomas Malory’s Linguistic Shift in the Bedtricks of Le Morte Darthur
Get PDFThis thesis explores the linguistic and narrative transformation of bedtrick scenes in Sir Thomas Malory’s Le Morte Darthur compared to their depictions in the Old French Vulgate. It investigates how Malory modifies the original sources’ romanticized portrayals of rape and deceit, shifting toward a rhetoric that aligns with his contemporary concerns, particularly regarding legitimacy and moral order. By analyzing three key bedtrick scenes – Uther and Igraine, Arthur and Morgause, and Lancelot and Elaine – this study reveals how Malory downplays the enigma and emotional justifications present in the Vulgate, opting instead for a sparser, more chronicle-like tone. This approach, as observed by scholars such as Corinne Saunders and Kathryn Gravdal, reflects a deliberate effort to sterilize and rationalize sexual violence, distancing it from the romantic glamorization typical of courtly romance. The analysis underscores that while the Vulgate blurs the boundaries between love and coercion, Malory reconfigures these narratives to serve his cultural and moral agenda, ultimately obscuring the reality of deceit, violence, and rape. The thesis concludes that despite their stylistic differences, both Le Morte Darthur and the Old French Vulgate participate in a broader literary tradition that seeks to rationalize and contain the disruptive potential of sexual transgression within the frameworks of their respective societies
The effects of front-of-package warning labels on consumer attitudes and purchase intentions toward reformulated food products
Get PDFTo address the growing concern of non-communicable diseases and obesity, the Government of Canada will introduce a mandatory front-of-package (FOP) warning label for prepackaged foods “high in” saturated fats, sugars, and sodium. This intervention, set to be implemented in January 2026, aims to nudge consumers into making better-informed and healthier food choices. Through a series of three experimental studies, this research seeks to examine: (1) the influence of the proposed FOP label on consumer attitudes and purchase intentions toward reformulated products; (2) the mediating effects of perceived healthiness and perceived tastiness; and (3) the moderating roles of goal salience (health vs. indulgence) and product type (healthy vs. unhealthy). Study 1 found that warning labels led to declining consumers’ attitudes and purchase intentions toward reformulated products, with perceived healthiness mediating the effect. Study 2 revealed that participants with a prominent health goal exhibited less favorable attitudes toward the reformulated product with (vs. without) the warning label, while those with an indulgent goal showed no difference in attitudes. Study 3 showed that participants demonstrated heightened sensitivity to the warning label when displayed on healthy (vs. unhealthy) food products. This thesis contributes to the literature on consumer behavior in the food industry, offering insights into the dynamics between health and indulgence within the framework of warning labels and product reformulation. The findings hold managerial implications for new product development, packaging, and communication strategies, and can help inform governments and policymakers about the effectiveness of warning labels for reformulated products
Design and Verification of Optomechanical Actuation for Tunable Filters
No full textThe exponential growth of network traffic, driven by applications such as IoT, 4K TVs, and cloud computing, demands an immediate need for enhancing network capacity. Optical fiber networks, fundamental to this enhancement, must handle massive data volumes efficiently. In order to address this, maximizing the existing optical network's capacity through dynamic networks and wavelength reconfigurability in wavelength division multiplexing (WDM) systems has emerged as a practical solution. Central to this approach are tunable optical filters (TOFs) and tunable lasers, which enable the selection and transmission of multiple data channels over a single fiber. In this work, a TOF with a 30 nm bandwidth (1535-1565 nm) across 5 channels, each with 30 nm tunability and 7.5 nm channel spacing, was selected to optimize performance within the C-band, known for its low attenuation and wide usage in optical communications. Silicon photonics, with its promise of smaller, faster, and more efficient photonic circuits, has advanced significantly, offering practical applications in integrated optical signal processors, external cavity lasers, and WDM systems. However, the mechanical tuning of photonic devices, crucial for dynamic network reconfigurability, requires robust actuators capable of delivering high force, broad travel range, and minimal footprint. This thesis investigates the design and optimization of MEMS actuators, specifically electrostatic comb drives with triangular finger configurations, to meet these stringent requirements. This research delves into various electrostatic actuator designs with a focus on optimizing force intensity, travel range, and spatial efficiency on Silicon-on-Insulator (SOI) platforms. It aims to enhance the performance of actuators for stiff, monolithic slab waveguides in silicon photonics, which requires high-force actuators due to their inherent mechanical rigidity. The study also explores the potential of angled electrode designs to improve microsystem actuator performance, providing a balanced approach to force intensity and travel range within a compact footprint. Additionally, the research addresses fabrication discrepancies and develops models to compensate for these variations, ensuring the reliability and consistency of manufactured devices. This thesis also covers the design and simulation of photonic components for a multichannel mechanically tunable optical filter and the external cavity of a diode laser. It involves the design of a concave diffraction grating (CDG) in Littrow mode, specifically tailored for compatibility with mechanical tuning. The study includes analytical modeling and validation through finite difference time domain (FDTD) and finite element method (FEM) with multiphysics simulations. The research further investigates the multichannel performance in various configurations and explores mechanical designs to reduce stiffness. Finally, the modeling and simulation of the designed multichannel optical filter are validated through experimental work, including selecting appropriate fabrication technology, performing scanning electron microscope (SEM) measurements, and testing the device using an edge-coupling setup. The fabrication of the optical filter configuration on Silicon-on-Insulator (SOI) platforms has been demonstrated, and the process has been considered in a way to facilitate easy post-processing for creating tunable optical filters. Overall, this thesis contributes to advancing silicon photonics by demonstrating that creating tunable filters on Silicon-on-Insulator (SOI) platforms is possible by providing innovative solutions for the mechanical tuning of photonic devices. These advancements, enhance the functionality, flexibility, and efficiency of optical communication systems
Chemical and electrochemical analysis of aqueous redox flow battery electrolytes
No full textRedox flow batteries (RFBs) are emerging as a robust solution for large-scale energy storage, crucial for meeting the growing demand for sustainable power sources. While conventional RFBs typically utilize metallic based electrolytes, aqueous organic redox flow batteries (AORFBs) are gaining momentum as a cost-effective and safer alternative. This study delves into the electrochemical characteristics of key redox couples, 1,2-dihydroxyanthraquinone (1,2-DHAQ) and ferrocyanide (FeCN) to assess their potential for AORFB applications. Employing cyclic voltammetry (CV) and rotating disk electrode (RDE) voltammetry, we evaluated their redox reversibility, pH stability, kinetic parameters, and diffusion coefficients. This study reveals that both 1,2-DHAQ and FeCN exhibit quasi-reversible redox behavior, with peak separations (ΔE) of 58 mV and 100 mV, respectively. The half-wave potentials (E1/2) versus an Ag/AgCl reference of 0.864 V for 1,2-DHAQ and 0.253 V for FeCN indicate their suitability as negolyte and posolyte, respectively. In addition, 1,2-DHAQ’s redox behavior is influenced by pH due to proton-coupled electron transfer (PCET) mechanisms, while FeCN’s redox potential remains stable across pH variations due to its coordination with cyanide ligands. Kinetic analyses showed that 1,2-DHAQ has a charge transfer coefficient (α) of 0.23 and a rate constant (k0) of 1.05 × 10⁻³ cm/s, compared to FeCN’s α of 0.7 and k0 of 1.2 × 10⁻³ cm/s. Diffusion coefficients, determined via the Randles-Sevcik equation, were 1.78 × 10⁻⁶ cm²/s for 1,2-DHAQ and 2.17 × 10⁻⁶ cm²/s for FeCN.
1,2-DHAQ electrolytes were cycled in both symmetric and full cell configurations and found to decrease in capacity over time. To understand the source of this capacity fade, aliquots of cycled electrolyte were analysed via CV and UV-Vis spectroscopy to track the concentration of 1,2-DHAQ over time. Although capacity fade was always associated with a decrease in 1,2-DHAQ concentration, the amounts were not consistently in quantitative agreement, suggesting that more work needs to be done to identify and quantify degradation mechanisms. Lastly, a combination of CV and open-circuit voltage (OCV) measurements were used to estimate the state of charge (SOC) of electrolytes. This method was found to be effective for quasi-reversible systems like 1,2-DHAQ and FeCN but less accurate for irreversible systems like the V4+/V5+ redox couple. The study also validated the use of a pseudo-reference electrode, which simplifies experimental setup and reduces contamination risks, thus enhancing the reliability of SOC measurements
Address to a few issues in thermoplastic composites made by automated fiber placement: Defect identification, temperature measurement, and model for tape width increase
No full textThe in-situ consolidation of thermoplastic composites (TPCs) via automated fiber placement (AFP) represents a promising method for manufacturing advanced high-performance structures, utilizing the distinct advantages of thermoplastics and enabling out-of-autoclave production. However, the widespread adoption of this method is impeded by various, poorly understood defects that can undermine structural integrity. This thesis begins by systematically identifying and characterizing these defects, providing a foundation for future research on their potential impact on mechanical performance. Subsequently, a novel experimental technique is introduced to monitor the temperature of incoming tape during in-situ consolidation, which is the most critical parameter controlling defect development. This is followed by an in-depth theoretical and experimental analysis of the transverse deformation of incoming tape under the compaction roller, a phenomenon that controls the size of gaps and overlaps as key defects in AFP-manufactured composite parts. The insights from this work can deepen the understanding of how specific process and material parameters directly influence the final tape width, paving the way for developing practical strategies to optimize AFP processes to manufacture higher-quality thermoplastic composites free of gaps or overlaps, and minimized dependent defects, e.g. fiber/ply waviness
To Taste a Portal: Contending with Cuban Colonial-era Histories through María Magdalena Campos-Pons’ Alchemy of the Soul, Elixir for the Spirits
Get PDFThis thesis examines María Magdalena Campos-Pons’ multisensory exhibition Alchemy of the Soul, Elixir for the Spirits (2015) as a means of resuscitating and preserving immaterial histories belonging to enslaved and indentured plantation labourers. The exhibition features immersive blown-glass installations, experimental soundscapes, an opportunity to taste freshly cut sugarcane and the faint yet persistent smell of rum wafting through the exhibition space. This thesis argues for the significance of the multisensory elements cultivating a space to contend with Cuban gastropoetics. I argue that access points to these immaterial archives are palpable and simultaneously protected through these multisensory elements. The multisensory details offer a possibility to experience the constellation of histories belonging to indentured and enslaved workers unified through Cuban sugarcane production. Alchemy of the Soul, Elixir for the Spirits does not scrutinize histories concerning plantation labourers—instead, Campos-Pons renders these histories opaque by privileging and questioning the role of the alimentary
tract and encourages the revitalization of cogent immaterial archives. Building from theoretical foundations laid by Jenny Sharpe, Édouard Glissant, Tao Leigh Goffe, Fernando Ortiz, Okwui Enwezor, and Parama Roy, this thesis advocates for the importance of Campos-Pons’ multisensory exhibition in the resuscitation and preservation of colonial-era histories degraded to immaterial status
Reappraising Beliefs About Losing Control: An Experimental Investigation
Get PDFBackground and Objectives: Beliefs about losing control over one’s thoughts, emotions, behaviours, and/or bodily functions have been shown to cause obsessive-compulsive symptoms. The cognitive model of obsessive-compulsive disorder (OCD) suggests that catastrophic misappraisals of intrusions will lessen if underlying maladaptive beliefs are effectively reduced.
The primary aim of this study was to experimentally investigate whether preexisting appraisals about losing control could be reduced by reappraising a previous perceived loss of control.
Methods: A sample of (n = 52) undergraduate participants underwent either a brief cognitive intervention or a control memory task. Appraisals about losing control and anxiety were measured before and after the manipulation.
Results: In comparison to participants in the control condition, participants in the experimental condition reported a significantly greater reduction in appraisals about having lost control (F(1, 50) = 10.79, p = .002, ηp2 = .18) and about losing control in the future (F(1, 50) = 7.82, p = .007, ηp2 = .14) but not anxiety F(1, 50) = .81, p = .37, ηp2 = .02).
Limitations: The absence of an impact on anxiety may be attributed to an underpowered sample size or the lack of a more robust intervention.
Conclusions: Results suggest that pre-existing beliefs about losing control can be reduced via a brief cognitive reappraisal-based intervention. Findings are discussed with respect to clinical and
phenomenological implications
Nioclás Tóibín: Déise Song Keeper
Get PDFThis thesis consists of ethnomusicological study of the life and legacy of Déise traditional sean-nós singer Nioclás Tóibín (1928-1994). Born in the Gaeltacht community of Na Déise, in Co. Waterford, Ireland, Tóibín was raised in the traditional Irish-language song culture of his community. He grew up in the first decade of Irish independence during a time of nation building. The new national identity, born from the cultural revivalism of the late 19th century, placed great symbolic importance on the promotion of the Irish-language and Gaelic culture, including music. During this time, there were also new technologies being introduced in Irish society, such as recording and broadcasting technology. Because of these factors, Tóibín’s musical career broke new ground for traditional singers from Gaeltacht na nDéise. Through interactions with revivalist institutions, such as Oireachtas na Gaeilge, and the promotion of Irish language music through recording labels like Gael-Linn and radio broadcast stations like Raidió na Gaeltachta, his voice became synonymous with the singing tradition of his community.
Using ethnographic interviews, this text is informed by the voices of sean-nós singers from Gaeltacht na nDéise who knew Tóibín, were influenced by him, and have experienced the unfolding of his legacy within their community since his death. This text considers the meaning of his memory, as well as how the success brought to Tóibín’s musical career through the interaction with recording and broadcasting technology has affected the sean-nós singing of Gaeltacht na nDéise
Navigating Change in the Era of Disruptive Technology: Insights from Hybrid Cross-Sector Partnerships in Digitalization
No full textThis dissertation contributes to advancing a comprehensive framework for understanding the socio-materiality of fields and their dynamics. The first article explores the co-evolutionary dynamics between technology trajectories and institutional pathways, emphasizing the temporal dimension of change, specifically the pace of change. In the second article, the focus shifts to the de-settlement of cross-sector partnerships within a technology field, highlighting how framing contests at the project level, coupled with partnership-level strategies, paradoxically escalate tensions in pathways and lead to cross sector partnership de-settlement. The third paper delves into the co-evolution of place trajectories by examining the interplay between imagination and emotional work. Collectively, these papers address two fundamental dimensions of change: materiality, represented by technology (papers 1 and 2), and place (paper 3), and temporality, specifically the pace of change (paper 1). While the first paper offers a framework for understanding the pace of change through the co-evolution of technology and institutional logic in field emergence and transformation stages, the second and third papers zoom in on the change trajectory during the decline/de-settlement stage, an area less studied in institutional change scholarship. These papers complement each other by focusing on temporality and dynamics while also considering durability and stability in the change process