Global ETD Search
Search theses and dissertations gathered from participating repositories worldwide. Every result links back to the library that holds it. No account is needed.
Results
Showing 1 to 18 of 18 for “"4603 Computer vision and multimedia computation"”.
-
Deep learning based stereo matching on a small dataset
Deep learning (DL) has been used in many computer vision tasks including stereo matching. However, DL is data hungry, and a large number of highly accurate real-world training images for stereo matching is too expensive to acquire in practice. The majority of studies rely on large simulated …
-
Multiview Registration of 3D Point Cloud with Feature-based Refinement
… including scene reconstruction, robotics and autonomous. In current studies, multiview registration has often been decomposed into two stages: pairwise registration and synchronization, where each one is considered as an independent task. In this thesis, we investigate a pipeline that can …
-
Automatic Detection of Neurological Disorders using Brain Signal Data
… disorders is a major challenge for public health and health care systems in Australia and around the world. Currently, there is no reliable way of identifying disorders from brain signal data automatically, quickly, and accurately. Electroencephalography (EEG) is a powerful and popular technique …
-
Deep learning-based skin cancer diagnosis
… samples under a microscope are time intensive and subjective, with notable variability between and within individual dermatologist evaluations. The advent of deep learning and advances in computational hardware have led to the development of computer-aided diagnosis (CAD) systems aimed at …
-
Automated Functional Assessment of the Fetal Heart in 4D Echocardiography using Deep Learning
… of the leading causes of death among newborns and studies have shown that early detection during the fetal stages reduces mortality and morbidity rates. Functional cardiac assessment is essential for early detection of such diseases and evaluation of in-utero fetal health. The aim of this …
-
Automated Detection and Monitoring of Vegetation Through Deep Learning
… but also for the development of sustainable and liveable cities. It is undeniable that human activities are altering the vegetation landscape, with harmful implications for the climate. As a result, autonomous detection, health evaluation, and continual monitoring of the plants are required …
-
Analysis of Cellular and Subcellular Morphology using Machine Learning in Microscopy Images
… based on machine learning models are data-driven and efficient and help to avoid subjective outcomes. However, the efficacy of these models is highly dependent on the feature description along with the amount and nature of the training data. This thesis presents three studies of automated …
-
The Artful Eye: Exploring Visual Engagement with Artworks in Different Contexts
… important to investigate the context-dependent and context-independent aspects of aesthetic experience. While some studies have shown that artworks in the museum elicit a higher visual engagement than when presented on a screen, others reported divergent findings. This thesis suggests that such …
-
Deep Learning Methods for Muscle Analysis From Magnetic Resonance Images
… imaging offers noninvasive tools for muscle size and structure analysis, typically necessitating pixel/voxel-wise annotations. Segmentation frameworks utilising deep learning offer a data-driven approach to automate the annotation process efficiently and accurately. However, their performance is …
-
From GANs to Diffusion Models: A Progressive Approach to Synthetic Data Domain Adaptation for Remote Sensing Change Detection
… annotated datasets, which are labor-intensive and expensive to create. While synthetic data offers a scalable alternative, a significant domain gap between synthetic and real imagery hinders the generalization of models trained upon it. This thesis addresses this critical challenge by proposing …
-
Automated Quantification and Epidemiological Characterisation of Regional MRI-Visible Perivascular Spaces in Brain Ageing
… growing burden of stroke, cognitive impairment, and dementia. Perivascular spaces (PVS) are increasingly recognised as MRI-visible markers of cerebral small vessel disease (cSVD), but their interpretation is limited by challenges in quantification, regional heterogeneity, and limited longitudinal …
-
Explainable Bias Mitigation in Transformer-Based Medical Image Analysis
… are often referred to as sensitive attributes. Vision transformers (ViT) have emerged as powerful deep learning models in medical image analysis, demonstrating outstanding performance across a wide range of tasks. These models are designed to learn very complex patterns in the given data, making …
-
Multimodal Retinal Deep Learning for Systemic Disease Detection with a Focus on Asthma
… disease with heterogeneous presentation and limited objective screening tools outside respiratory testing. Current detection relies on spirometry and bronchial challenge tests that are effort-dependent, operator-sensitive, and often impractical in primary care or paediatrics, contributing …
-
Symbolic Neural Reasoning for Visual Question Answering
… of Visual Question Answering (VQA) systems and the latest Large Vision and Language Models (LVMs) represents a significant intersection in multimodal intelligence, which is a fundamental aspect of human intelligence. Humans excel at using multimodal data, combining vision, language, and …
-
Deep Learning Based Prognosis and Explainability for Breast Cancer
… necessitating advancements in diagnostic and prognostic methodologies. This research aims to develop and evaluate novel deep learning frameworks that leverage the rich information embedded in histopathology images, combined with molecular and clinical data, to improve prognostic accuracy …
-
Apple Flower Density and Phenology Estimation for Chemical Thinning and Data-Centric Analysis of On-tree Fruit Detection
… marketable fruit. Estimating the flower counts and phenology is important for chemical thinning decisions. Farmers generally inspect flowers in randomly sampled trees within the orchard, which is time-consuming, labour-intensive, and unreliable. Existing algorithms for estimating flower density …
-
Swimming in Data: Using AI and distance sampling to improve the analysis of remote underwater videos in estimating fish abundance
… that are typically analysed using biased and underpowered metrics, complicating the inference required for marine fisheries management. This thesis makes three contributions, to improve the utility of Remote Underwater Videos (RUVs). Firstly, animal movement behaviour is exploited to …
-
Learning Visual Attention and Gaze Communication Behaviours with Applications to Autism Spectrum Disorder Diagnosis
Humans employ a complex combination of verbal and non-verbal communication to convey information and express intentions. The limited ability to effectively communicate hinders knowledge acquisition and even results in mental health disorders such as autism spectrum disorders (ASD). There are …