Honours and Masters project

Cybersickness (nausea, disorientation) due to exposure to virtual environments has long been a problem in virtual reality (VR) and has been shown to reduce the effectiveness of VR environments. It usually occurs due to mismatches between visual and vestibular (motion) cues, for example, when moving through a 3D environment using a joystick, which does not yield correct motion cues. There are several approaches to reducing cybersickness in VR, most notably, reducing the field of view ("tunneling") during motion, or discrete motion ("snapping" movement and rotation).

The performance of deep neural models rely on large amounts of labeled data, however, most data remain unlabeled in the real world scenario. While annotating data is expensive and time consuming, active learning seeks to choose the most appropriate and worthwhile data for human annotation. It is noticed that humans give labels to some specific data with some labeling reasons or rationales,  which are often existing in the data.  The goal of this research is to develop effective deep active learning techniques with rationales.

Deep learning has achieved ground-breaking performance in many vision tasks in the recent years. The objective of this project is to apply the state-of-the-art deep learning based image classification/detection networks such as ResNet or Faster RCNN for classifying CT or X-Ray images.

Studies using computed tomography (CT), particularly high-resolution CT and quantitative CT have become a crucial non-invasive method for examining airway thickness and the structural changes known as airway remodeling in chronic respiratory diseases such as asthma. Bronchial thermoplasty is a treatment option for patients with severe asthma and works by applying heat energy to reduce the amount of excess airway smooth muscle, which is often abnormally thickened in patients with asthma.

In vitro fertilisation (IVF) is a process of fertilisation where an egg is combined with sperm outside the female body, in vitro ("in glass"). The process involves monitoring and stimulating a person's ovulatory process, removing an ovum or ova (egg or eggs) from their ovaries and letting sperm fertilise them in a culture medium in a laboratory. After the fertilised egg undergoes embryo culture for 2–6 days, it is implanted in the same or another person's uterus, with the intention of establishing a successful pregnancy [1].

This project will involve benchmarking state of the art methods for time series classification on the new MONSTER benchmark datasets [1, 2, 3].  Currently almost all benchmarking in time series classification is performed on the (almost all very small) datasets in the UCR and UEA archives.  This is particularly unsuitable for deep learning models which are low bias models and ideally trained using large quantities of data.  The "true" performance of current deep learning methods for time series classification is unknown outside of the UCR/UEA datasets.  Most deep learning models for times…

Description:

Magnetic Resonance Imaging (MRI) stands as a cornerstone in medical imaging, providing non-invasive, high-resolution images of the human body's internal structures.  Brain tumor segmentation from MRI scans is essential for precise diagnosis and treatment planning. MRI provides detailed views of brain structures and abnormalities, but challenges like image noise, contrast imperfections and tumor variations can make segmentation difficult.

Traumatic brain injury (TBI) is an injury to the brain caused by an external force from incidents such as motor vehicle crashes, falls, assault or sports collisions. Almost seventy million individuals globally are estimated to suffer from TBI per annum [1], deeming it a major public health concern which is estimated to cost the global economy approximately $US400 billion annually [2]. Early identification of severe TBI with proper assessment and treatment lowers the risk of secondary injury and subsequent long-term disability and subsequent costs.

Deepfakes detection deals with machine learning methods, which detect if an image/video/audio sample is manipulated with a generative AI software. In recent years, deepfakes have been increasingly used for malicious purposes, including financial fraud, misinformation campaigns, identity theft, and cyber harassment. The ability to generate highly realistic synthetic content poses a serious threat to digital security, privacy, and trust in media. This project will develop methods for detecting deepfakes.

People are continuously receiving unsolicited emails where phishers impersonate legitimate organisations or trusted sender to harvest victim credentials. The rapid advance of AI boosts recent automatic detection of phishing attempts but also provides hackers with the opportunities to build increasingly sophisticated phishing tactics to bypass the filter. While attackers leverage social engineering to exploit human weakness, human skills can be a powerful component in cyber defence such as cognitive function and professional judgment.