Procter & Gamble (United Kingdom)
Open Access Institutional Repository at Robert Gordon UniversityNot a member yet
9563 research outputs found
Sort by
M3OT: a multi-drone multi-modality dataset for multi-object tracking. [Dataset]
No full textUnmanned Aerial Vehicles (UAVs) have been widely used in many fields, such as smart city construction, environmental monitoring, power system inspection and traffic management. Compared with static cameras, UAV imaging can cover more expansive areas and more various viewpoints, which facilitates the detection and tracking of moving targets in rapidly changing environments. As shown in Fig. 1, UAV-captured images exhibit unique characteristics, such as dynamic viewpoint changes and irregular object movements, which significantly increase the complexity of tracking tasks. To address these challenges, the research community has developed a variety of UAV-based datasets. Many of these are dedicated to object detection, such as CARPK5 for vehicle counting, AU-AIR for low-altitude multi-modal detection, the high-altitude thermal dataset HIT-UAV, and TinyPerson for small-scale person detection. While these benchmarks are crucial for advancing object detection, they are generally designed for single-frame analysis. Consequently, they often lack the continuous, inter-frame identity annotations essential for developing and evaluating algorithms for multi-object tracking, which is a distinct and more complex task. Multiple object tracking (MOT) aims to accurately track multiple objects across consecutive image frames, achieving by assigning each object a unique identity (ID) while maintaining the consistency of object identities. Most existing MOT approaches adhere to the Detection-Based Tracking (DBT) framework. As illustrated in Fig. 2, the DBT method begins by employing an object detection algorithm (e.g., Faster R-CNN or YOLO) to identify the location and category of objects in each frame. After that, a matching algorithm is used to associate these detected objects with existing tracklets, thereby ensuring continuity in tracking. New objects that appear are assigned new tracklets, while tracklets for objects that are consistently detected are updated over successive frames. Conversely, tracklets are terminated when an object is no longer detected or exits the field of view. With the rapid development of computer vision, several MOT datasets have been proposed for performance evaluation, such as MOTChallenge and DukeMTMC. However, most of these datasets are collected from surveillance videos, which inherently possess certain limitations. For example, the annotated bounding box areas in these videos are predominantly larger than 1024 pixels, and the tracklets tend to follow linear motion patterns
From pixels to perspectives: exploring perceptions of representation in character design.
Get PDFThis study presents findings about perceptions of how women characters are introduced in a sample of six best-selling games. Our analysis adds to ongoing conversations about representation in games and media, encouraging both designers and researchers to reflect on how perceptions and emotions related to representation shape interactions from the perspectives of gamers and non-gamers alike. We analysed 198 responses to a survey that asked participants questions about how they felt about short videos of character introductions and captured open-ended responses about the roles these characters occupied. We used descriptive and comparative statistics to identify basic trends in closed-ended survey responses amongst participants. We then identified, applied, and iterated on emergent descriptive qualitative codes designed around participant's open-ended responses. We identified a series of themes regarding cultural context, perceived characterization and character roles, and time and dialogue. Our findings build on previous literature about character design and women's representation in games with analysis based on community perceptions and provide considerations for anyone designing characters and their introductions or working with diverse groups to make the medium and industry more inclusive
From learning to action: exploring barriers and strategies for embedding lessons learned through a social learning lens.
Get PDFThe purpose of this study is to explore the barriers to implementing lessons learned in project-based organisations and how they can be effectively captured and embedded to create knowledge that drives project success and organisational performance. The study adopted a qualitative research design to explore the perceptions and experiences of project practitioners through the lens of social learning theory (SLT). Semi-structured interviews were conducted to uncover individual project practitioners’ views and opinions about lessons learned in the project environment to facilitate project success. The findings indicate that effective application of lessons learned and knowledge sharing significantly enhances project success by fostering a culture of collective learning and continuous improvement. The key barriers identified are misinterpretation of lessons learned, lack of mentoring, knowledge hiding, non-participation from relevant stakeholders and lack of a collaborative environment. The research therefore indicates that organisations should formulate clear guidelines, implement robust knowledge repositories and standardise lesson-learned processes to foster uniform understanding within and across projects. Organisations should also invest in training programs and promote mentorship and top management support to foster a culture of learning and accountability. This paper presents a unique avenue to understanding lessons learned in the context of project-based learning. It highlights barriers and strategies across dimensions and explores them through a participatory and social lens. By using an SLT theoretical framework, the study provides a fresh perspective on how knowledge can be shared and utilised within project organisations. It bridges gaps in literature and provides actionable insights, making it highly valuable to both academic and practical fields
Near miss: behavior and effect.
Get PDFNear miss plays an important role in providing insights and are also leading indicators of potential hazards within a system, one could argue that they are attributed to safety behaviours, human or organisational factors, and are underreported. The study will focus on exploring the behavioural factors that contribute to Near miss, explore the perceptions of relevant stakeholders on the factors influencing reporting behaviour, identify challenges of the reporting system, explore and understand the influence of Near miss reporting on continuous improvement. Data is collected using qualitative research method (semi structured interviews) to ensure the interviewee is given the opportunity to give insights and specifics to subject. A set of questions were prepared prior to the interview as a framework to guide the discussion. The research was carried out using purposive sampling which is a non-random sampling method to achieve a good correlation between the collected samples and the objective of the work with a ratio of 1: 10 to have a variable perspective presenting an analysis of interviews with 14 participants (PT), exploring their varied understandings of near miss - which ranged from "prevented accidents" to "potential harm with no injury." The study highlights near miss reporting as essential for preventing more significant incidents and promoting workplace safety awareness. Participants emphasized the need for clear reporting processes, strong leadership involvement, and continuous feedback loops to improve safety practices. Key barriers identified include fear of judgment, a lack of understanding about the benefits of reporting, and communication challenges. To address these issues, the study recommends improving training, simplifying reporting through technology, and fostering a non-punitive safety culture that encourages open communication
Embedding an entrepreneurial mindset in postgraduate engineering courses for real world societal benefits.
Get PDFThis engaging talk aims to inspire HE educators to embed an entrepreneurial mindset in engineering education, aligned with the Advance HE's Framework for Enterprise and Entrepreneurship Education. It highlights a collaborative initiative at Robert Gordon University, where two new MSc modules were developed. These modules encourage postgraduate students to address global challenges related to the UN Sustainable Development Goals through innovative engineering-based solutions. Utilising a flipped classroom approach, the modules foster critical discussion, project development, and presentation skills, and serve as a pipeline to the university’s Startup Accelerator Programme, demonstrating the tangible benefits of integrating entrepreneurship in engineering curricula
Resistance training beyond momentary failure: the effects of past-failure partials versus initial partials on calf muscle hypertrophy among a resistance-trained cohort.
Get PDFThis study compared calf training with initial partial repetitions versus full range-of-motion (ROM) repetitions followed by past-failure partials on gastrocnemius hypertrophy. Twenty-three participants (men: n = 16 and women: n = 7) performed four sets of unilateral Smith machine calf raises to momentary failure twice a week for 8 weeks. One leg was trained using initial partials to their individualized maximum dorsiflexion ROM. The contralateral leg was trained with a full ROM and continued with past-failure partials after failure in peak plantarflexion. Medial gastrocnemius muscle thickness was measured with ultrasonography both baseline and postintervention. A Bayesian framework was used to estimate the average treatment effect (ATE) using credible intervals and Bayes factors (BFs). The ATE posterior distribution indicated a greater increase in muscle hypertrophy for the initial partial condition (0.40 [95% CrI: −0.06 to 0.85 mm]; p (> 0) = 0.958), with a BF of 1.2 suggesting "anecdotal" evidence in favor of an effect. Within-condition analyses using standardized mean difference estimates indicated that the interventions were likely to produce medium to large improvements. These findings suggest that both initial partials and past-failure partials are viable strategies for achieving gastrocnemius hypertrophy. Although the average change favored initial partials, the estimated difference was uncertain, and the Bayes factor provided only anecdotal support for a differential effect
A 10-year review of periconceptual folic acid supplementation in women with epilepsy taking antiseizure medications.
Get PDFEpidemiological studies have reported that women with epilepsy who are taking antiseizure medications have an increased risk of Neural Tube Defects. Periconceptual folic acid supplementation potentially prevents two-thirds of cases. International guidelines recommend that women at increased risk of a pregnancy complicated by a Neural Tube Defect who could become pregnant should start high-dose (5 mg daily) oral folic acid at least three months before conceiving. The purpose of the study was to examine supplementation in women taking antiseizure medications who delivered a baby weighing >499 g during the ten years 2013–2022 in a large maternity hospital. The hospital's computerized database contains standardized maternal clinical and sociodemographic details which were entered at the first antenatal visit in the obstetric records. The data on all women with epilepsy taking antiseizure medications was anonymized before coding. In the ten years, 75,869 women delivered a baby weighing >499 g. Of the deliveries, 632 (0.83%) were to women with epilepsy. Of these, 250 (0.33%) were taking antiseizure medications when they presented for antenatal care. The most frequently prescribed medications were lamotrigine n = 98 (33.8%) and levetiracetam n = 89 (30.7%). Monotherapy was prescribed in 211 (84.4%) women and polytherapy in 39 (15.6%). Three (1.2%) women took no folic acid before or after conception and 59 (23.6%) only took it after conception. Of the 188 (75.2%) who took folic acid before conception, 164 (65.6%) took high-dose 5 mg and 24 (9.6%) took low-dose 0.4 mg. No maternal characteristics were associated with taking high-dose folic acid before conception. Compliance with the national guidelines in the 16 women taking valproate was 30.8% compared with 76.0% for the other 234 women taking medications (p < 0.03). Compliance in the 211 women receiving monotherapy was 72.5% compared with 25.6% in the 39 women receiving polytherapy (p < 0.03). Two-thirds of women taking antiseizure medications complied with national guidelines on high-dose periconceptual folic acid supplementation. A particular concern is the suboptimal compliance in women prescribed valproate and polytherapy, which are two cohorts at higher risk of a neural tube defect. These findings need to be communicated to women with epilepsy in their reproductive years and their doctors. Of 75,869 women delivered over 10 years, 250 (0.33%) were on medications for epilepsy at presentation 34.4% of women taking medications did not start high dose folic acid before conception 23.6% of women did not start folic acid until after conception Compliance with recommendations was lowest in women prescribed valproate or polytherapy No maternal characteristics were identified which were associated with suboptimal compliance
Investigations into the physical layer of underwater networks to enhance channel capacity and performance.
Get PDFThe Underwater Internet of Things (UIoT) needs development to enable usage of modern network applications underwater, particularly regarding sustainable high capacities to enable multimedia transmissions that can monitor industrial or environmental processes. This thesis investigates the nature of the physical layer of underwater networks and investigates how it can be developed to pragmatically increase their channel capacity and general performance according to common networking metrics, so that data can be effectively and efficiently transmitted from the ocean floor to the surface enabling passive harvesting of data. It investigates replacing low-capacity ad-hoc acoustic networks with a smart, co-operative hybrid-architecture that uses high data rate wireless visible light communication to transmit information through the underwater environment towards the floating sink nodes on the surface. This link is supported by optic fibre and artificial intelligence to connect source nodes to each other with smart functionality. This allows for source nodes to share information that enables them to co-ordinate successful transmission of data to the sink whilst avoiding common channel disturbances that would otherwise result in a failed transmission. The thesis shows the investigations and methodology utilised to suggest this as a valid architecture. Firstly, this work investigated how the acoustic network could be optimised through MATLAB simulation using known models from the literature. It was identified that through use of floating repeater nodes at periodic distances to form an optimised multi-hop network, or at least a single link with a signal regeneration mechanism in place, acoustic communication can take place with a prospective maximum capacity increase, from 31.2kbit/sec in a modern modem to 2Mbit/sec, that has the potential to carry a diverse range of multimedia. However, despite the capacity increase, it still lacks the potential to carry high channel capacity multimedia such as video and high-quality images with reasonable quality, firstly due to the relatively low channel capacity but also due to the high latencies associated with acoustics. Thus, it was decided to follow up these results with an investigation into a wireless optical communication-based architecture that has the potential to increase the channel capacity beyond what found in acoustic communications, so that these applications are more feasible. It was found that, according to MATLAB simulations and modelling, an underwater wireless optical network can be developed that increases the possible maximum capacity from the 2Mbit/sec from to 10Mbit/sec and reduces latency, from 188ms to 0.8ms, so that these multimedia applications become more viable. However, it was determined during simulation that this technology is not going to achieve this consistently because of local noise, a variable transmission range due to fluctuating organic matter concentrations and vulnerability to objects. Similarly, it was decided that further investigations were necessary to find methods to enable this consistency, thus, it was decided to investigate the combination of two architectures working in co-ordination to deliver successful outcomes. It was decided that a smart wireless acoustic-optical hybrid network would be compared to a strategically combined smart optic fibre-wireless optical network for capacity and performance comparisons through simulation in MATLAB. It was found that the latter has the potential to successfully maintain consistently high capacity of 10Mbit/sec and low delay jitter of 1ms whereas the former jittered significantly in capacity and latency as the acoustic link acted as a significant bottleneck of 5Mbit/sec as a theoretical maximum, resulting in capacity jitter of 5Mb/sec and delay jitter of 67.7ms. Thus, it would add unpredictability to the network and adversely affect transmission data for applications with high-capacity demands. However, this paradigm with higher capacities is entirely dependent on being able to determine when the forementioned noise, organic matter and objects are likely to interfere with operational certainty. Finally, to research an appropriate manner of mode switching from optic to the deferral mechanism according to the local environment, it was decided to establish an early concept of "cognitive optics" which aims to render a network cognitive of the underwater wireless optical channel. It aimed to achieve this through utilisation of an artificial intelligence algorithm that can carry out decision making based on sensed data on noise, local particulate presence and objects that will allow for the transmission method to switch according to local conditions surrounding the node. The investigation utilised a synthetic dataset developed through the utilisation of established models and known oceanic boundaries to train a series of accurate machine learning classifiers, specifically random forest, neural network and support vector machines. In addition, the potential of a fuzzy logic controller was investigated for this purpose. It was found that either fuzzy logic or machine learning classifiers could accurately decide whether to transmit or co-operate with another source node to complete transmission. In terms of machine learning algorithm performance, the experiment showed that a small neural network could achieve this fastest and generally most accurately whilst being least likely to use visible light when it should co-operate with another node to maintain the investigated reliable capacity performance of 10Mbit/sec. For future works, the aim is to create an underwater network that can operate autonomously utilising intelligent modem agents to self-optimise their placement, modulation and power according to circumstances in the network. This would expand on the theories investigated in this thesis by looking at mechanisms that will enable the network to self-optimise in the face of environmental and network conditions in complex manners using artificial intelligence, avoiding the need for costly interventions. The work is concluded with the evaluation that an autonomous underwater multimode network is the best solution to realising the Underwater Internet of Things, most crucially though, that there is potential beyond the dominant acoustic/optical hybrid network topology for other mechanisms to be used such as the investigated co-operative, smart, optical network to transmit data effectively through the water column to the source. There are many future works that can be investigated based on the results, such as further automation and cognitive functionality of the nodes and spatial multiplexing concepts to further raise capacity for transmission in the network
An exploration of the information literacy of Irish theatre actors.
Get PDFInformation literacy (IL) from a sociocultural perspective is understood as being the ability to understand and interpret a given environment, and to use information to perform increasingly complex socially embodied tasks therein towards expertise in any field. Empirical IL studies of diverse working populations are influential in demonstrating and espousing contextually situated and tacitly embodied knowledge as the meaningful foundation upon which professional knowledge and practice are built. To date however, there have been no IL studies exploring the explicitly embodied work of actors. The aim of this research is to investigate the IL of Irish actors working in the theatre. The established rhetorical framework for examining working populations is Workplace Information Literacy (WIL). However, this study suggests that the terminology and consequent perspective of IL, rather than WIL, is appropriate in approaching actors, as their workplaces are unpredictable, differentiated and transitory by design. As a result, actors' practice is flexible and adaptive towards the many contexts in which they work, thus a study of a single workplace would omit inclusion of this diversity and limit empirical insight. Furthermore, as freelance workers, actors' IL necessarily extends beyond formal workplaces of rehearsal and performance, into varied formal and semiformal professional contexts, and informal socially networked environments. Therefore, IL offers a simplified, holistic and inclusive lens with which to consider the complex diversity of actors' professional lives and practice. This study examines how influential WIL theory and research exploring embodied and sociocultural workplace learning applies to the context of actors' experience of formal practice. It critically examines the use of IL as a conceptual framework for a study of actors across the diverse formal and informal contexts of their freelance careers by means of a holistic exploration of actors' experience of professional life. An inductive, constructivist philosophical orientation, employing a qualitative, multimethodological approach and phenomenological design was followed. Interviews with a sample of twelve participants, each interviewed three times, using a different interview technique for each meeting, was conducted. Open-ended constructivist, and collective, constructionist questions about the experience of professional life and practice were asked, avoiding questions explicitly pertaining to IL, so that participants could express their IL-related experience authentically. Interview to the Double, the second interview technique, challenged participants to describe their individualised practice in subjective detail. The cumulative and durational nature of the interview cycle allowed participants to reflexively offer insight into the data gathering process and their contribution to the research. Data was analysed following thematic coding using NVivo software, and findings validated through a process of self-examination supported by heuristic inquiry. The findings show that actors working in the theatre experience IL in three interdependant contexts: in Theatre Practice, within the Theatre Industry, and as a Theatre Artist. In theatre practice, IL is instance-dependent upon engagement with each company uniquely assembled to develop a production. The rehearsal process requires a sophisticated and robust negotiation of the social environment towards the realisation of a shared value held in common that is the unifying vision for the production. This is a socially embodied process in which the body is a practical workplace tool and source of finite, carefully managed and deployed energy, as well as the somatic site of attentive play towards presence in the moment of performance. Within the theatre industry, beyond formal theatre practice, IL is practiced by actors as freelance artists across the various landscapes of the industry, in interactions with their agents, auditioning as part of the casting process, staying informed, and self-motivated professional development. The theatre industry is experienced as a community of practice within which mentors, peers and friends inform practice and decision-making, and professional engagement occurs across the socially networked landscapes of the arts. As a theatre artist, across the multifarious, unpredictable diversity of these formal, informal and semi-formal landscapes, actors develop a personal belief system within which to intuit meaning, validation and value based on an experiential, reflexive passion for and commitment to both the theatre-making process, and life as an artist. The belief system is robust and individualised, adaptive and responsive to new contexts, and supports spontaneous, transformational lifelong learning, through which a complex, resilient and self-sustaining identity as an artist may emerge, independent of external authorisation. Based on these findings, a foundational model of actors' IL is presented. This thesis builds upon emergent empirical and theoretical work in the field of IL espousing the significance of the sociocultural approach to illuminate diverse working populations and that foreground embodiment and tacit, experiential knowing as being key to the universality of IL in all contexts. This study is unique in approaching the population of actors in order to illuminate IL, and offers a model of such that may be applied, tested and developed across diverse acting, arts and freelance contexts. The model of actors’ IL emphasises the essential connective tissue of a constructive, individualised belief system grounded in a commitment to praxis to support learning and identity across contexts. The methodologically distinctive contribution of this study is that the researcher too is an actor steeped in the culture, language, norms and conventions of the sample and population, allowing for a rare interdisciplinary study and the novel application of heuristic inquiry, lending a consequent transparency and validity to the thesis
LKVHAN: multi-scale large kernel vertical-horizontal attention network for hyperspectral image classification.
Get PDFAmong deep learning-based hyperspectral image (HSI) classification models, convolutional neural networks (CNNs), Transformers, Mamba, and large kernel CNNs (LKCNNs) models have been widely explored for HSI classification. Nonetheless, these models suffer from several challenges: for example, 1) CNNs have a weak learning ability in capturing global information between land covers, due to their limited receptive field derived from small kernel convolutions; 2) Transformers face quadratic computational complexity introduced by their self-attention mechanisms; and 3) LKCNNs require further enhancement in extracting global features, owing to the insufficient size of their receptive fields. To tackle these limitations, we propose a novel multi-scale large kernel vertical-horizontal attention network (LKVHAN) for HSI classification. The proposed LKVHAN consists of a 1×1 convolution module and a multi-scale large kernel vertical-horizontal attention-based convolution (MSLKVHAC). The 1×1 convolution module is designed to facilitate band reduction, noise suppression, and spectral feature learning. Furthermore, the MSLKVHAC, leveraging a large vertical kernel size of 17×1 and a large horizontal kernel size of 1×17, extracts both local and global spatial features by incorporating a vertical attention-based convolution module (VACM) and a horizontal attention-based convolution module (HACM). Extensive experimental results demonstrate that the proposed LKVHAN significantly outperforms ten state-of-the-art approaches across four widely used HSI datasets
Data sources:
Access Repository Dashboard
Do you manage Open Access Institutional Repository at Robert Gordon University? Access insider analytics, issue reports and manage access to outputs from your repository in the CORE Repository Dashboard!