National University of Ireland, Maynooth
MURAL - Maynooth University Research Archive LibraryNot a member yet
18981 research outputs found
Sort by
Learned Descriptors for Scalable and Efficient Visual Place Recognition
Get PDFVisual Place Recognition (VPR) is a task in computer vision that involves matching
images of an environment to previously visited locations, enabling systems to identify
and recognise places based on visual information. VPR has emerged as a widely studied
topic in computer vision and mobile robotics, driven by its applications in autonomous
navigation, image retrieval, and loop closure detection. Over the past decade, the field
has witnessed significant progress, fuelled by improvements in camera hardware, the
proliferation of mobile devices, and the growing availability of public image datasets.
Researchers have increasingly utilised deep learning techniques to tackle the challenges
of VPR, particularly those related to appearance changes and varying viewpoints that
traditional descriptors struggled to address.
Despite these advancements, several interconnected challenges hinder the deployment
of reliable and scalable VPR systems in automotive applications. Utilising largescale
sequential datasets poses significant difficulties due to diverse recording conventions,
redundant visual content, and limited viewpoint variance, complicating training
processes for deep learning. Additionally, efficiently categorising scenes without explicit
object identification introduces considerable computational and methodological
complexities. Furthermore, VPR systems face challenges related to scalability, primarily
due to the computational demands associated with rapid retrieval of images
for localisation along extensive trajectories spanning several kilometres. This thesis
specifically addresses these critical challenges.
First, we introduce OdoViz, a comprehensive and unified framework designed
for efficient dataset exploration, visualisation, analysis, curation, and preparation of
bespoke training data from heterogeneous datasets. OdoViz streamlines the creation of
standardised, tailored datasets essential for robust VPR model training.
Secondly, we elaborate on the development of robust learned image descriptors
utilising large sequential datasets. We introduce a novel discretisation approach that
segments trajectories into visually similar regions, facilitating efficient online sampling
of triplets for contrastive learning. We present a detailed training regime involving
tailored data subsets, a modified architecture, and a custom loss function for stable
contrastive training, optimised to generate robust learned image representations.
Thirdly, we propose an efficient scene categorisation method leveraging Variational
Autoencoders (VAEs). Our approach encodes images into compact, disentangled
latent spaces without explicit object recognition, enabling rapid categorisation into
urban, rural, and suburban contexts. This method achieves exceptional computational
efficiency, with inference times under 100μs, making it suitable for use as a pretext task
in real-time automotive applications.
Finally, addressing the scalability concerns, we introduce a hierarchical framework
utilising learned global descriptors to facilitate rapid retrieval over extensive distances
while maintaining robust localisation performance. Through extensive experimentation,
we identify continuity and distinctiveness as key properties of effective global descriptors
for scalable hierarchical mapping, and propose a systematic method to quantify and
compare these characteristics across various descriptor types. Our VAE-based scene descriptors
achieve up to 9.5x speedup on the longest evaluated track, St Lucia (17.6km),
while maintaining the same recall performance over longer trajectories, demonstrating
their effectiveness in hierarchical localisation.
Together, these contributions address the identified VPR challenges, laying the
groundwork for scalable and efficient VPR systems leveraging learned representations,
suited for deployment in diverse real-world automotive environments
Examining the Role of the State in Doctoral Education: A Review of the ‘National Strategy for Higher Education to 2030’
No full textGlobalisation positioned doctoral graduates as ‘human capital’, equipped with the knowledge and skills to strengthen research and innovation to gain a competitive advantage globally (Bernstein, BL et al., 2014; Nerad, 2020; Nerad et al., 2022; Nerad & Trzyna, 2008). This study aimed to investigate how the Irish government recontextualised doctoral education, addressing the problems identified in the ‘National Strategy for Higher Education to 2030’ (‘Hunt Report’) and assess the instruments used to govern its implementation. In addition, the study sought to critically evaluate the outputs of the Hunt Report in terms of human capital and changes in activities and behaviours within the Triple Helix of Government-University-Industry. The objective was to explore the impact of the Hunt Report on the transformation of doctoral education in Ireland.
Using Colebatch’s Policy Framework, the study first conducted a critical discourse analysis of the Hunt Report, assessing the impact of global and national policies between 2005 and 2011 and the technologies of governance outlined in the Hunt Report. Secondly, over ten years after the Hunt Report, the study critically examined the outputs and changes in behaviours on two key themes: governance and human capital. This analysis drew on quantitative data obtained from OECD and Central Statistics Office (CSO) and the Department of Public Expenditure & Reform Databank databases, a review of seven Irish universities’ websites and the System Performance Framework (2018-2020) and Mission-Based Performance Compacts (2018-2021)
Arctic sea ice thickness prediction using machine learning: a long short-term memory model
Get PDFThis paper introduces and details the development of a Long Short-Term Memory (LSTM)
model designed to predict Arctic ice thickness, serving as a decision-making tool for maritime
navigation. By forecasting ice conditions accurately, the model aims to support safer and
more efficient shipping through Arctic waters. The primary objective is to equip shipping
companies and decision-makers with a reliable method for estimating ice thickness in the
Arctic. This will enable them to assess the level of risk due to ice and make informed decisions
regarding vessel navigation, icebreaker assistance, and optimal sailing speeds. We utilized
historical ice thickness data from the Copernicus database, covering the period from 1991 to
2019. This dataset was collected and preprocessed to train and validate the LSTM predictive
model for accurate ice thickness forecasting. The developed LSTM model demonstrated a
high level of accuracy in predicting future ice thickness. Experiments indicated that using
daily datasets, the model could forecast daily ice thickness up to 30 days ahead. With monthly
datasets, it successfully predicted ice thickness up to six months in advance, with the monthly
data generally yielding better performance. In practical terms, this predictive model offers a
valuable tool for shipping companies exploring Arctic routes, which can reduce the distance
between Asia and Europe by 40%. By providing accurate ice thickness forecasts, the model
assists in compliance with the International Maritime Organization’s Polar Code and the Polar
Operational Limit Assessment Risk Indexing System. This enhances navigation safety and
efficiency in Arctic waters, allowing ships to determine the necessity of icebreaker assistance
and optimal speeds, ultimately leading to significant cost savings and risk mitigation in the
shipping industr
A Monthly Snow and Sleet Series for the Greater Dublin Area 1867–2024
Get PDFThis paper details the compilation of data and application of quality assurance procedures for constructing a 157‐year snow and sleet series for the Greater Dublin Area, Ireland. Snowfall is particularly sensitive to climate variability in temperate regions, and long‐term records are essential for understanding changes in winter weather extremes over time. The dataset integrates observations from six sites and provides a regional snow and sleet frequency dataset at monthly, seasonal (October–May) and annual resolutions. Data sources include archived meteorological records, digitised station logs and synoptic weather reports. A brief analysis offers insights into long‐term snowfall climatology in the Greater Dublin region from 1867 to 2024, revealing substantial interannual and decadal variability, as well as notable reductions in snow frequency in recent decades. This dataset provides a valuable baseline for assessing historical trends in snowfall and contributes to broader efforts in climate reconstruction and climate change impact studies in Ireland and beyond
Sunitinib-derived Pt(IV) complexes display enhanced anticancer activity against renal cell carcinoma compared to conventional platinum chemotherapy
Get PDFHalf of all cancer treatments worldwide involve the use of Pt chemotherapeutics but despite
their wide clinical usage, Pt drugs have severe disadvantages, including cell toxicity. Sunitinib
is a FDA approved Tyrosine Kinase Inhibitor which selectively targets renal cell carcinoma due
to the overexpression of its receptors such as vascular endothelial growth factor receptor
(VEGFR) and platelet derived growth factor receptor (PDGFR). Here, a family of three Pt(IV)
prodrugs, based on the clinically approved cisplatin, oxalilplatin and carboplatin, bearing
sunitinib-derived axial ligands have been developed with the aim to overcome healthy cell
toxicity. This study highlights the first Pt(IV) complexes targeting renal carcinoma tumours
overexpressing VEGFR. Conjugation of the sunitinib-based ligand was shown not to jeopardize
its kinase inhibitory activity. In vitro cytotoxicity proved the cisplatin prodrug derivative to be
36 times more active to cisplatin chemotherapy and 3D spheroids assays reinforced this
superior activity. The cisplatin-based prodrug was tested in vivo against renal carcinoma
xenografts, revealing exceptional superior activity to cisplatin control. This work
demonstratesthe excellent potential of Pt(IV)-Sunitinib conjugates for the treatment of Renal
cell carcinoma, as they display far enhanced tumour reduction and lower systemic toxicity
when compared to cisplatin chemotherapy
Impact of ischemic lesion on sleep related connectivity in the sensorimotor cortex
Get PDFIschemic events can cause cell death and tissue loss, leading to the impairment of neural circuitry by disconnection of its neural substrates. However, the highly plastic properties of the nervous system can provide recovery by boosting circuital redundancies or triggering functional adaptation/repurposing of closely related networks. In this context, understanding how ischemic brain lesions reorganize circuits directly or indirectly connected to the injury site is crucial for developing therapeutic approaches, particularly neuroprostheses based on neurostimulation for brain-rewiring. Furthermore, it is also fundamental to consider the sleep–wake cycle in such an inquiry, considering its well-established role as bearer of key mechanisms of neuroplasticity. This study aimed to investigate how an ischemic lesion in the rat’s primary motor cortex affects the connectivity of areas involved in the sensorimotor loop, specifically the premotor cortex (RFA) and the primary somatosensory cortex (S1), during sleep. We analyzed Local Field Potentials recorded during slow-wave sleep in rats with and without ischemic lesions. Functional connectivity and cross-frequency interactions were quantified using Phase Locking Value (PLV) and Phase-Amplitude Coupling (PAC) analyses, respectively. Our findings revealed a marked increase in PAC 7 days after the lesion, followed by a partial return toward baseline levels at 14 days post-lesion. These results suggest a transient reorganization of network dynamics associated with early recovery processes. The observed changes provide insights into spontaneous post-stroke plasticity during sleep and identify potential electrophysiological biomarkers of recovery. Our findings may contribute to the design of sleep-integrated neurostimulation strategies to promote motor rehabilitation after stroke
Investigating MSC Immunomodulation in the Obese Microenvironment
No full textThe proportion of people living with obesity is constantly on the rise, which makes it an important factor to consider in any clinical setting. The obese microenvironment introduces a range of pro-inflammatory factors and other molecules that can affect immune cells and treatments. Mesenchymal stromal cells (MSCs) are a valuable asset for the treatment of inflammatory conditions due to their immunomodulatory and their tissue regenerative abilities. MSCs are able to communicate with surrounding immune cells and respond to signals from the microenvironment. However, this makes MSCs vulnerable to perturbations in that microenvironment. This thesis investigates how certain prominent factors from the obese microenvironment, the free fatty acids oleate and palmitate and the adipokine leptin, affect MSC phenotype and function. Differences between T cell suppression and macrophage suppression by palmitate exposed MSCs were discovered, and the underlying mechanism for how palmitate affects macrophage suppression by MSCs was further elucidated. In addition to the direct effects of these factors on MSCs, we were also interested in the concept of innate immune training of macrophages by factors elevated in the obese microenvironment, and how MSCs might be able to interfere with this training. Innate immune training can lead to a hyperinflammatory activation of macrophages, which, while protective in many cases of infectious disease, has the potential to result in a more severe outcome of chronic autoimmune disease. MSCs, which are powerful suppressors of macrophage inflammation, have already been shown to suppress house dust mite-mediated training in our lab. Building on this research, a training protocol was established and pathways through which MSCs may be inhibiting training were investigated. Overall, this research revealed avenues to be explored in MSC-macrophage interactions, highlighting the potential for the use of MSCs in patients with inflammatory conditions who are living with obesity, and identifying pathways through which innate immune training could be controlled
“Here Lads, this is the story!”. An exploration into access practitioner knowledge and community engagement within Irish higher education
Get PDFThis research explores higher education access practitioner knowledge, investigating the relationship between access practice and community engagement. The research delves into the world of HEI access practice, it looks at how access can be enhanced drawing on the knowledge from professional access practitioners and communities that experience disadvantage.
Access practitioners have been working within Higher Educational Institutions in Ireland since the late 1990s. Despite equity of access being a strategic priority of the Higher Education Authority and policy and funding commitments by the Government to address educational inequalities, there remains inequalities in our society that have deep levels of educational disadvantage. As an access practitioner for over two decades, this research explores with access colleagues and community participants if and how community engagement practices could enhance access practices.
This research stands within a critical tradition, interested in questions regarding equality and power. Using qualitative research methods with community participants and access practitioners, the reality of the access role is depicted, alongside the untapped access opportunities that exist within communities. In-depth interviews with practitioners and community workshops using participatory research methods (photovoice) were undertaken. Research participants explored themes relating to educational disadvantage, access to higher education and community engagement.
The research found that access practice at institutional level is significantly impacted by neoliberal government policy and new managerial practice. The empirical evidence from this research points to difficult working conditions, inappropriate institutional positioning, and pressures on time for access practitioners, all of which limits real meaningful engagement with communities that are under-represented in higher education. Access practitioners are working with limited resources, and with time specific funding streams, which have negative consequences for community engagement. Communities have been on the receiving end of this hurried approach, resulting in very few opportunities for meaningful, collaborative and respectful engagements, where HEIs and communities can
together, as equals, address issues relating to educational disadvantage. New principles for access and community engagement for access professionals are presented
Report: The ERC DANCING Final Conference
Get PDFThis report gives an account of the DANCING Final Conference, which was held on 19 and 20 June 2025. The conference presented the main findings and results of the European Research Council-funded project ‘Protecting the Right to Culture of Persons with Disabilities and Enhancing Cultural Diversity through European Union Law: Exploring New Paths – DANCING’, led by Professor Delia Ferri (Professor of Law at the School of Law and Criminology of Maynooth University). The DANCING Final Conference aimed to reflect on the key achievements of the project, but also to explore future directions for research, policy, and practice related to European Union (EU) disability law and the right to cultural participation of persons with disabilities.
The conference brought together academics, policymakers, persons with disabilities, artists, and other stakeholders. Over two days, participants engaged in thematic panels and roundtables. The DANCING project, running from 1 September 2020 to 31 August 2025, pursued three main objectives: to identify barriers and facilitators to cultural participation by persons with disabilities (experiential), to explore how EU law can promote disability rights and cultural diversity (normative), and to offer a new theorisation of cultural diversity in EU law (theoretical). All three objectives were discussed throughout the conference. The first day focused on the research findings and academic contributions of the project, while the second day turned to policy implications and ‘Tools for Change’, including the launch of a Policy Brief – comprising recommendations for European Union and national policymakers and a Toolkit for cultural stakeholders.
Notably, the conference was opened by a video message from Professor Adam Bodnar, Minister of Justice of Poland, on behalf of the Polish Presidency of the Council of the European Union (1 Jan 2025 – 30 Jun 2025). Two keynote speeches – respectively from Professor Bruno De Witte (Emeritus Professor at Maastricht University and Member of the Advisory Board of the DANCING Project) and Ms Inmaculada Placencia Porrero (Member of the UN Committee on the Rights of Persons with Disabilities and Senior Disability Expert at the European Commission) – marked the two days. A keynote dialogue led by Professor Mark Priestley (Emeritus Professor at Leeds University and Member of the Advisory Board of the DANCING Project) concluded the second day of the conference. The DANCING Final Conference also included two side events that marked the artistic collaborations deployed during the course of the project. The first was the unveiling of the artwork ‘Odisseo-Ulysses’ created by artist Tiziano Pantano. The second was the premiere of the documentary ‘Steps to Change: Following the DANCING Project 2020 – 2025’ produced by Feenish Production Ltd. and directed by James Kelly.
Besides discussing the conference content and its most significant moments, the report outlines key organisational aspects with a focus on how accessibility was planned and delivered throughout the event, in line with the UN Convention on the Rights of Persons with Disabilities (CRPD) and DANCING’s own accessibility strategy. Strategic planning was vital to ensure venues were accessible to wheelchair users, but also suitable for people with a range of disabilities. Accessibility measures further included provision of Irish Sign Language (ISL) interpretation throughout, alternative formats (Braille, large print, digital soft copy) of the programme and relevant documentation of the conference, as well as the embedding of accessibility features in the documentary and the artwork.
On the whole, the DANCING Final Conference showcased the cross-cutting nature of the interdisciplinary research conducted, but also highlighted the project’s commitment to bridging academic research with practice and to ensuring that persons with disabilities are active participants in shaping inclusive cultural spaces across Europe