3789 research outputs found
Sort by
An improved semi-supervised learning framework for Image semantic segmentation
Full text linkTraditional supervised learning methods depend heavily on labeled data, which is both costly and
time-intensive to acquire. Self-supervised learning approaches present a promising alternative to
supervised learning, enabling the utilization of unlabeled data. Thus, this research aims to build
an advanced semi-supervised semantic segmentation model that strikes a balance between selfsupervised and fully supervised paradigms for visual perception applications in an autonomous
driving environment.
In this direction, the thesis is structured into three distinct phases, beginning with self-supervised
image classification and progressing toward bi-level image segmentation, ultimately culminating
in the development of an advanced semantic segmentation model. Initially, this research employs
a simple contrastive learning framework (SimCLR) to classify medical images, specifically focusing on monkeypox diagnosis from skin lesion images, while integrating a federated learning (FL)
framework to ensure data privacy. Monkeypox classification is a simple binary classification task
and the dataset found for this problem, in this thesis, is very manageable on the computational resources that were available at the onset of this research. It paved the way to grasp non-supervised
learning basics and explore how they differ from traditional supervised learning methods.
The subsequent phase involves the development of an efficient convolutional neural network
(CNN) with an attention mechanism, applied to the bi-level segmentation task of road pavement
crack detection. Similar to the Monkeypox classification, this is also a binary classification task,
but at pixel-level, i.e., it is a two-way semantic segmentation problem. Hence, the number of
samples found in the relevant datasets is once again manageable on the computational resources
available during the research. [...
Planning for battery electric buses charging in transit system
No full textWith the growing focus on sustainable transportation, Battery Electric Buses (BEBs)
have emerged as a viable solution. BEBs have received significant recognition as an
environmentally conscious and sustainable means of transportation. In many cases,
transitioning from a conventional diesel-fueled transit system to a fully electric one is
essential. Designing an effective strategy, which encompasses placing charging sites and
implementing proper charging mechanisms, is crucial to ensuring efficient and consistent
charging of BEBs in an electrified public transit system. However, the challenge intensifies
when the transit planner aims to maintain a consistent daily service timetable.
The research endeavours to tackle this challenge by formulating efficient charging strate-
gies and methodologies for infrastructure planning. This thesis outlines a four-step
approach for transit system planners to attain optimal solutions, encompassing worst-case
energy consumption calculation, off-service charging site placement, off-service (overnight)
charging mechanism, on-route charging planning, and finding the number required BEBs
and integration of them to fully electric transit system. Four methods are designed
for use in planning: the Constrained Greedy Clustering (CGC) algorithm, the Priority
Charging Mechanism (PCM), the Constraint Affinity Clustering Algorithm (CACA), and
timetable tuning. A case study based on a real-world Thunder Bay, ON transit system
validates the proposed methodologies and assesses their effectiveness in improving the
overall performance of the BEB fleet. Results demonstrate significant improvements in
operational efficiency, cost reduction, and environmental sustainability by implementing
the proposed charging infrastructure optimization strategies.
The findings of this research contribute to the advancement of sustainable transportation
by providing practical insights and solutions to the challenges associated with BEB
charging infrastructure design and optimization
Exploring the role of climate emotion tools in climate science learning with youth and Indigenous climate activists from Mexico
Full text linkOne of the biggest problems faced by activists, youth, Indigenous people and other people at the
forefront of climate impacts and the climate movement is inaction by governments and the
majority of the world’s population. In response, there is a growing trend of introducing climate
change education in different learning environments, from schools to universities to informal
education in activism spaces, resulting in a wide range of climate-related emotions in learners.
This study seeks to bridge the gap between research on climate education and climate emotions
by exploring the possibilities and impacts of combining climate science education with peer-topeer tools that aim to enhance climate science learning and support the emotional resilience of
youth and Indigenous climate activists. Deeply influenced by community-based participatory
action research, this qualitative study consisted of semi-structured interviews with seven
volunteer participants following a six-class online climate science course that was developed in
collaboration with 12 potential course participants from two activist networks. The emotions
tools were integrated into the classes, and participants found that these tools greatly benefit their
lives, learning processes, and mental health, suggesting the importance of embedding socioemotional learning environments in both educational and activism contexts to foster community
building and proactive engagement
The design of fast-transient cap-less low-dropout voltage regulators
Full text linkThis thesis provides a theoretical and experimental study of cap-less LDO regulators
for high speed applications. The three different architectures used in LDO designs are re-
viewed in detail along with their advantages and disadvantages. Theoretical analysis of each
architecture is covered along with a review of state of the art designs. The thesis presents
two cap-less designs. The first design uses a dual loop architecture to enable fast transient
response and high current capability of which the current loop offers fast transient while the
voltage loop provides regulation. The second proposed LDO utilizing a hybrid architecture
of which the digital part introduces a fast transient approximation algorithm demonstrating
significant speed improvements over traditional algorithms. The design is optimized for low
clock frequency applications and high output current. Both LDO designs are manufactured
using the TSMC 180 nm technology and demonstrate both simulation and measurement
results
Development of an advanced thermal imaging-based human fall detection system
No full textFalls represent a significant risk to the elderly population, often leading to severe injuries or fatalities.
Automatic fall detection systems (FDS) are critical for mitigating these risks; however,
existing solutions, despite reporting accuracies in controlled environments, often fail to generalize
to real-world conditions. This performance gap stems from limitations in existing datasets,
overfitted models, and a lack of standardization. To address these challenges, this thesis presents
a comprehensive framework for fall detection, leveraging privacy-preserving thermal imaging to
develop deployable, real-world solutions.
The research is conducted in three progressive phases. The first phase explores a novel hybrid
architecture that combines supervised and unsupervised learning paradigms through a stacking ensemble
of 3D Convolutional Neural Networks (3D CNNs) and Autoencoders (AEs). This hybrid
approach demonstrates significant performance improvements on constrained datasets, highlighting
its potential in scenarios where fully supervised methods fall short. Ablation studies validate
the architecture’s utility while underscoring the critical need for a more robust dataset to achieve
true generalizability.
In the second phase, the thesis introduces Thermal Fall 66 (TF-66), the most diverse and comprehensive
thermal fall detection dataset to date. Designed to address the limitations identified
in Phase 1, TF-66 encompasses varied environments, participant demographics, and fall scenarios.
Accompanied by targeted subsets and flexible data generators, TF-66 serves as a benchmark
for meaningful comparisons and standardized evaluations, advancing the field toward real-world
applicability.
The third phase synthesizes insights from the hybrid approach and TF-66 dataset to refine a
supervised 3D CNN model. Enhanced with innovative features such as optical flow integration and
attention mechanisms, this model achieves state-of-the-art performance on TF-66 and the widely
used Thermal Simulated Fall (TSF) dataset. By bridging the gap between lab-optimized systems
and real-world demands, this final phase establishes a transformative approach to fall detection,
redefining the state of fall detection research, with a focus on generalizable systems that can operate
in real-time. The findings provide a clear path for developing accurate, privacy-preserving, and
scalable fall detection systems, ultimately aiming to enhance the safety and save lives of seniors
worldwide
Housing crisis and international students' health in Canada: a case study at Lakehead University
Full text linkThis research seeks to examine the housing experiences of international students, utilizing
Lakehead University as a case study site to ascertain the potential effects of the prevailing housing
crisis and different forms of housing on international students' physical and mental well-being.
This study employed a qualitative approach, combining semi-structured interviews and
document analysis. Interviews with international students offered in-depth insights into housing
experiences, while document analysis review added context on policies at Lakehead University.
Purposive sampling identified 15 participants, including students and staff. Using NVivo, a thematic
analysis revealed key themes, triangulating findings from both the interviews and institutional
documents to capture the structural housing determinants.
The study found that international students preferred on-campus housing initially, transitioning
to off-campus options later due to affordability concerns. Common challenges included high costs,
maintenance issues, and landlord-tenant relations. Good housing conditions positively impacted wellbeing
and integration, while poor conditions negatively affected mental health. Coping strategies
involved online platforms, social networks, and university support, leading to recommendations for
expanded on-campus housing and affordable policy changes.
The housing crisis in Canada significantly impacts international students, who face financial
strain, substandard living conditions, and inadequate institutional support. Addressing this requires a
comprehensive approach from governments, universities, and housing providers to create affordable,
safe, and accessible options tailored to international students' needs. Resolving the housing challenges
is essential for international students' well-being, academic success, and integration into Canadian
society
Assessing graphite precipitation mechanisms of the Albany graphite deposit
No full textGraphite is recognized as a 'critical raw material' due to its strategic importance for diverse
industries (e.g., steel, automobile, clean technologies). The Albany graphite deposit, located west
of Hearst in Thunder Bay, is a fluid-derived, igneous-hosted deposit. The morphology of the
Albany deposit is unique because of fine-grained graphite occurring within two large breccia pipes.
Albany graphite deposit is located in the south of the Nagagami Alkali Complex, north of the
Gravel River fault. The objective of this study is to determine the mineralogy and alteration
assemblages (non-weathering) using mineral compositions and textural associations and evaluate
the alteration assemblages to determine whether they play any role in graphite precipitation.
In the Albany deposit, graphite typically occurs as elongated, lath-shaped, and plate-like
crystals (<0.05 mm in width and ranges from 0.1 to 1.5 mm in length). Crystals are typically
characterized as randomly oriented and showing both intergranular and intragranular textures.
Graphite is mainly distributed along grain boundaries in clast components (intragranular), whereas
it shows intergranular texture in the matrix. The abundance of graphite in the matrix (<15%) is
higher than that in the clast (<10%). Three sub-groups of graphite, namely clast, matrix, and
alteration assemblages were examined. The matrix assemblage is comprised of lithic fragments
derived from the clasts. Clasts are mainly comprising varying proportions of graphite, plagioclase,
potassium feldspar, quartz, biotite, amphibole, and chlorite, while the matrix is mainly comprised
of equigranular graphite, plagioclase, potassium feldspar, quartz, and biotite. Hydrothermal
alteration phases present biotite, sericite, calcite, and chlorite. Biotite and sericite are determined
as pre-graphitization alterations due to their textural relationship with graphite. Calcite forms
partial and rim replacement of plagioclase in the matrix, while chlorite replaces biotite and alters
plagioclase along grain boundaries. Petrographic observations reveal that both calcite and chlorite
alterations are directly associated with graphite in the Albany deposit. Calcite and chlorite, which
constitute at the propylitic alteration assemblage, developed coevally with graphite. [...
Geochemistry and mineralizaton of the Lundmark Akow area, North Caribou Greenstone belt, Ontario
Full text linkThe Lundmark Akow area is located in the south-central porƟon of the South Rim assemblage of
the North Caribou Greenstone Belt. The mineralizaƟon consists of several base metal-bearing massive
sulphide horizons hosted in a sequence of garneƟferous staurolite mica schists in the southern porƟon of
the study area, to a northern host rock sequence dominated by volcanic and intrusive mafic to felsic
rocks. The igneous host rock porƟon of the study has been dated between 2973 to 2980 Ma through
zircon U/Pb geochronology. The meta-sedimentary sequence which hosts the mineralizaƟon is built
upon a basement of intrusive and volcanic rocks which formed in an oceanic plateau through plume
magmaƟsm before impinging upon a subducƟon zone. NegaƟve high field strength element (HFSE)
anomalies show that arc related magmaƟsm built upon the oceanic plateau. Sm/Nd isotope values from
the mafic to felsic volcanic and intrusive rocks show a spread in εNd from -1.53 to 3.07 suggesƟng that
melts were derived from both depleted mantle and plume sources, with some melts being contaminated
by an older crustal basement.
The characterisƟcs of the massive sulphide horizons, as well as the host meta-sedimentary rocks
are consistent with them having formed through distal VMS processes including hydrothermal
parƟculate fallout from buoyant plumes combined with pooling of dense sulphide-rich fluids in
topographic lows on the sea floor. The garnet-rich meta-sedimentary rocks show Fe and Mn enrichment
when normalized to immobile Al and Ti, consistent with addiƟon through hydrothermal plume
parƟculate processes. The characterisƟcs of the garnet layers suggest they formed through
metamorphism of a sediment derived from the intermixing of hydrothermal and terrigenous
parƟculates. The garnet composiƟons show prograde growth with a Mn- and Ca-rich core to a Fe- and
Mg-rich rim, supporƟng their formaƟon as a result of metamorphic condiƟons post exhalaƟve acƟvity. [...
Estimating DBH with iPad Pro LiDAR in Boreal Forests: methodological considerations and a case study in natural boreal forests
Full text linkDiameter at Breast Height (DBH) is the measure of the diameter of a tree stem 1.3
meters above the ground. DBH is a key variable measured in Forest Resource
Inventories (FRIs) and is traditionally measured manually, which is labour-intensive.
The 2020 Apple iPad Pro 12th Generation is a lightweight, consumer-level tablet with an
integrated LiDAR scanner with a maximum range of 5 m and a positional accuracy of
±1 cm. The overall objective was to examine the feasibility of estimating DBH in boreal
forests with iPad Pro LiDAR. A scoping study was conducted in a plantation forest
(48.37°N, 89.39°W) near Thunder Bay, ON, Canada, with the specific objective of
determining an optimal method for DBH estimation with the iPad Pro. Different
combinations of scanning methods (i.e., circular, figure-8, transect), numbers of stem
cross-sections (i.e., one or five), sizes of stem cross-sections (i.e., 4 or 10 cm), and curvefitting formulas (i.e., Pratt’s circle fit, Taubin’s circle fit, Taubin’s ellipse fit, Szpak’s
ellipse fit) were tested to identify the combination producing the most accurate
estimates of DBH. The optimal method was the circular scanning pattern with a single 4
cm cross-section and a combination of circle- and ellipse-fitting formulas (RMSE = 1.1
cm; 6.2%). The second specific objective was to determine the accuracy of DBH values
estimated with the optimal method in natural boreal forests. DBH was estimated for 133
trees on 15 sites in northern Ontario, Canada, representing a range of natural boreal
forest site conditions. A secondary objective was to determine if the tested stand- (i.e.,
species composition, age, density, understory density) or tree-level attributes (i.e.,
species, actual DBH) significantly impacted the accuracy of estimated DBH values. An
RMSE of 1.5 cm (8.6%) was achieved. Estimated DBH was within 1 cm of actual DBH
for 78 of 133 (59%) measured trees. Stand age had a large effect (> 0.15) on the accuracy
of estimated DBH values, while density, understory density, and actual DBH had
moderate effects (0.05-0.15). In both studies, Inertial Measurement Unit (IMU) and
positional accuracy errors with the iPad Pro LiDAR scanner limited the accuracy of
DBH estimates. Future studies should incorporate a greater number of natural boreal
forest sites to better understand the impacts of different stand and tree attributes on the
accuracy of estimated DBH values. Future studies should also compare the accuracy of
DBH values estimated from the iPad Pro and those estimated from traditional MLS and
TLS for the same sites to identify the trade-off between device cost, device size, and
accuracy. However, the scanning range of the device limits the variables that can be
estimated from LiDAR data, rendering it unsuitable for use in FRIs until the scanning
range is improved
A comparison of reports of fatal grizzly and black bear attacks in Canada
Full text linkThe purpose of this study is to determine if there has been an increase in the
number of fatalities caused by grizzly and black bear attacks in Canada from 1990 to
2023, as well as to determine if the attacks were predatory or defensive in nature. The
data used in this study was acquired from various news sources in Canada as well as the
United States to gain a better understanding of the fatalities and events leading up to the
attack. The information was separated into tables, and they were used to establish the
number of fatalities caused by grizzly and black bears, province, or territory the age and
gender of the victims, and the month which the attacks occurred in. Overall, the number
of fatalities has remained relatively similar over the years, there was a decrease in
fatalities cause by black bears in 2010 to 2019. During the years from 2020 to 2023,
almost half the number of fatalities as previous decades, likely because more people
were spending time outdoors due to the COVID-19 pandemic. Many reports lacked
specific information about the attacks and events leading up to the encounters, and
therefore it was difficult to determine if the attacks were predatory or defensive in
nature