Chalmers Open Digital Repository
Not a member yet
26247 research outputs found
Sort by
New tools for old news
No full textMany collections of digitized newspapers suffer from poor OCR quality, which impacts readability, information retrieval, and analysis of the material. Errors in OCR output can be reduced by applying machine translation models to “translate” it into a corrected version. Although transformer models show promising results in post-OCR correction and related tasks in other languages, they have not yet been used for correcting OCR errors in Swedish texts. This thesis presents a post-OCR correction model for Swedish 19th and 20th century newspapers based on the pre-trained transformer model ByT5. Three versions of the model were trained on different mixes of training data. The best model, which achieved a 37% reduction in CER, will be integrated in Språkbanken Text’s annotation pipeline Sparv
Prospects of Normalizing Flow-Based Differentiable Particle Filters in Target Modelling
No full textAccurate target modelling is essential for modern air surveillance, particularly in
the face of increasingly dynamic and unpredictable target behaviour, such as that
exhibited by hypersonic missiles and UAVs. Traditional tracking methods, including
Kalman Filters, Interacting Multiple Model (IMM) filters, and Particle Filters
(PF), rely heavily on known target dynamics and are often limited in adaptability.
This thesis explores the potential of Differentiable Particle Filters (DPFs), and
in particular Normalizing Flow-based DPFs (NF-DPFs), as a data-driven alternative
for modelling and tracking targets with unknown or highly non-linear dynamics.
To address these challenges, a NF-DPF was implemented for radar-based target
tracking. The filter was designed as a fully differentiable architecture that learns
both the proposal distribution and the system dynamics using conditional normalizing
flows based on the RealNVP framework. It incorporates optimal transport-based
differentiable resampling and was trained end-to-end using a combination of a blockwise
Evidence Lower Bound (ELBO) loss and the RMSE in position, velocity, and
acceleration. The training was conducted on target data generated from stochastic
differential equations (SDEs) simulating a range of motion patterns, including
turning and accelerating trajectories. The NF-DPF was benchmarked against a
bootstrap PF and a IMM filter across three experiments with increasing data complexity
and variations in filter state dimensionality.
Results showed that the NF-DPF provided competitive performance in low-dimensional
settings and under specific dynamic behaviours, particularly turning trajectories
that were not well captured by the IMM’s filters. However, it struggled to outperform
the IMM in more complex scenarios involving higher-dimensional state representations.
This performance gap was largely attributed to the increased difficulty
of learning high-dimensional transition and proposal distributions, combined with
computational limitations such as a restricted number of particles.
The results highlight both the potential and current limitations of NF-DPFs for
target modelling and contribute to the growing body of research exploring the integration
of flexible, learning-based models into particle filtering frameworks for
defence and surveillance applications
Transformer-Based Multi-Object Tracking of Football Players Using Pseudo-Labeling
No full textThis thesis investigates the problem of tracking football players in video sequences, with a focus on adapting modern multi-object tracking (MOT) methods to the specific challenges of football environments. The work is part of a broader effort to develop tools for analyzing football games using automated visual data. In this context, we utilize MOTRv2, a transformer-based tracking model originally designed for general-purpose MOT tasks, and apply it to the football domain, where challenges such as frequent occlusions, tight formations, and rapid movement are prevalent. To address the lack of annotated football-specific tracking data, we implement a pseudo-labeling framework that allows the model to be trained on unlabelled domain data in a semi-supervised fashion. This approach enables progressive refinement of the model through multiple training cycles on domain-specific content. Our results show that MOTRv2 can be adapted to the football setting and performs well in many scenarios, particularly in open-play segments with clear player separation. However, limitations remain, including decreased tracking stability in crowded scenes and occasional ID-switches due to overlapping motion patterns.
Overall, this work demonstrates the potential of transformer-based trackers in sports applications and highlights the benefits of self-supervised training when domainspecific data is scarce. The findings offer insights for future improvements in automated sports tracking systems
Non lonely architecture; an alternative approach for resilience to loneliness
No full textThe growing prominence of loneliness in public discourse underscores the
need for environments fostering social interaction. Stemming from unmet
needs for social connection, loneliness prompts questions about how
relationships are formed and the role of everyday spaces in building social
networks. This thesis thus investigates architecture’s potential to address
this challenge by facilitating relationship-building opportunities in local
contexts. Specifically, in Hammarkullen, Gothenburg.
Adopting a pragmatic context-sensitive approach,
mappings of existing conditions in the locality informed
gaps to in social opportunities. Leading to the proposed
intervention; a third place in the form of a resident centred
café, complementing the existing local social fabric.
The methodology is built on human-centered practices.
Recognizing that social engagement is subjective and
dependent on human perceptions and socio-spatial
relationships. This led to a mixed-methods approach
adopting from discourses on loneliness and co-creation
architecture. Key methods applied cover; Relational mapping
of social systems informing the project positioning,
involving traditional architectural methods, interviews
and the use of relationscapes.
Relationscapes were also applied in thematic mappings
of system dynamics. Further, destinations were identified
and evaluated to their social quality towards loneliness
using qualitative and quantitative methods concluding
a mapping of social opportunities in the locality. In the
design process, a participatory process was conducted
with a focus group, co-creating the café. Overall, the
methodology and theoretical framing are influenced by
fields of human geography, gerontology, and relational
architecture. The results shaped a conceptual narrative
of a resident-centered café. Developed through civil,
public, and grassroots collaboration. Wherein process,
the architect takes on a role of facilitating relationships in
addition to the traditional role of spatial expertise.
This thesis concludes that co-creation processes can
produce social values relevant to mitigating loneliness
while also fostering shared ownership of urban spaces
and strengthening local agency. Furthermore, it infers
that relational mappings of dynamic social systems can
serve as tools to decipher where and how interventions
can be most effective. Thus, contributing to the broader
discourse on loneliness, participatory architecture, and
citizens’ rights to shape their built environmen
Hands-On Algebra - An Experiential Learning Intervention in Middle School Using Physical and Virtual Manipulatives
No full textIs there any difference in knowledge gained from using a virtual object rather than
a physical in a learning session? Several studies show the benefit of using manipu-
latives for teaching. What is possible using technology changes everyday, including
what learning experiences can be created. This makes new research extra relevant
for this area of interest. Following an experiential learning intervention of two blocks
with pre- and post-tests using virtual and/or physical manipulatives that resulted in
a total of four sessions in six classes in Swedish middle school. The mean difference
between post- and pre-test was 8.1 percentage units with N=62. Analysing results
using a two-way and three-way ANOVA can not statistically confirm if the format of
only virtual, only physical or a combination of both impacts mathematical learning
differently. However, the format does affects other attributes such as engagement,
accessibility and customisability. Manipulatives can be a promising way to teach,
and further research with more classes is encouraged
Piecewise Diffusive Score-based Generative Model
No full textDiffusion-based generative models have achieved remarkable success across a range
of applications. However, their performance often degrades in the presence of imbalanced
data. To address this limitation, we introduce a novel generative modeling
framework that combines stochastic differential equation (SDE) driven by Brownian
noise with Poisson random measures. We also provide the explicit form of reverse
SDE, with corresponding loss function to train the neural network. Our experimental
results show that the model with Poisson jump outperforms the model without
jump across various toy datasets and a small subset of CIFAR-10
Intermediary Strategies for Supporting Companies with Policy Implementation Challenges A Multisectoral Study of EU ETS 2 and the Green Claims Directive
No full textDespite ambitious climate targets set by the EU and Sweden, a significant
implementation gap remains between policy goals and real-world actual outcomes. This
thesis investigates how an intermediary organisation, such as RISE Research Institute
of Sweden, can support companies in closing this gap with focus on two forthcoming
EU directives – the revised EU Emission Trading Scheme (EU ETS 2) and the Green
Claims directive (GCD) in Sweden. The study specifically aims at identifying
challenges that companies in the Swedish energy, manufacturing and building sector
face in relation to these directives and identifying strategies that intermediary
organisations can develop to address these.
A qualitative case study approach was used, involving 9 semi-structured interviews
within the different sectors and a workshop with stakeholders from RISE. Findings
were analysed using an analytical framework based on prior literature, that included
both a system-and actor-level dimension.
Regarding challenges companies faces, the findings showed a wide range of barriers at
the both the system-and actor-level. System-level challenges included policy risk- and
design issues, financial constraints, cultural resistance and infrastructural barriers.
Actor-level challenges related to limited company resources and organisational
structures specific to individual companies. To address these challenges, several
intermediary strategies that organisations like RISE can adopt were identified across
three dimensions: managing and regulating, capacity building and problem solving. At
the system-level, these were related to offering verification services and providing
clear, accessible information. At the actor-level, they involved delivering educational
support and consultancy services. While some challenges were addressed by multiple
intermediary strategies, particularly those related to infrastructure and company
resources, others related to culture and policy risk-and design remain largely
unaddressed.
The outcome of this study contributes to theory by offering empirical insight on
challenges and strategies related to EU ETS 2 and GCD in a Swedish setting. Some
findings aligned with previous literature while some were newly discovered,
highlighting new dimensions to the policy implementation gap. The multi-levelled
strategies identified may also be scalable, offering potential support for companies with
other climate policies. Ultimately, the thesis concludes that an intermediary
organisation like RISE can play an important role in helping companies comply with
the challenges identified in this study
Metanol som framtidens bränsle inom sjöfarten: Kan metanol möta framtidens miljökrav i enlighet med EU:s direktiv?
No full textSjöfartsindustrin står inför stora förändringar i val av bränsle för att nå de globala klimatmålen. Internationella sjöfartsorganisationen (IMO) har som mål att år 2050 uppnå nettonollutsläpp av växthusgaser. För att uppfylla dessa krav krävs alternativa bränslen, där metanol är ett av alternativen. Denna studie har undersökt både fossil och grön metanol för att bedöma dess potential inom sjöfarten. Studien fokuserar på tre huvudfrågeställningar:
För att besvara dessa frågor genomfördes semistrukturerade intervjuer med rederier, tekniska experter och bränsleleverantörer för att få en bredare bild av industrins syn på metanol som drivmedel.
Resultaten visar att fossil metanol inte är ett hållbart alternativ, då dess utsläpp vid produktion är högre än dagens konventionella bränslen. Däremot kan aktörer i Göteborgs Hamn anpassa sin infrastruktur för att hantera metanol inom ett år med mindre modifieringar. Intervjuer med rederier tyder på att det finns ett intresse för metanoldrivna fartyg, men den begränsade tillgången på grön metanol inom EU utgör en betydande utmaning för en övergång till detta bränsle
