Making sense from Images with Deep Learning

Closing Date

31st December 2021*

*unless filled earlier

The Research Project

Deep Learning recently lends itself extremely well to the research in computer vision domain where hierarchical structures of computational neurons can learn predictive features to effectively make predictive decisions. For example, in health care, deep learning is becoming also popular among medical imaging researchers who are looking for great tools to process a large number of images produced by scanners.

The impact of this to the society is potential and attract more and more attention from health care experts who have been looking for better methods to reduce the error rates in diagnosis. However, the most common deep learning models used for image processing are CNN-based which is a complex black-box consisting of millions of parameters that confused the experts of why the decisions are made. As a result, there is an increasing scepticism from those who do not want to use deep learning because of the lack of explainability.

In this research, the student will improve the transparency of deep neural networks to provide insights of the decision-making process. The topics of interest are (but not limited to):

  • Medical imaging (eye disease detection, knee pain prevention, etc.)
  • Visual reasoning, image captioning
  • The project is open to Australian (domestic) and international candidates
  • The PhD must be undertaken on a full-time basis
  • Honours degree/Master degree or hold equivalent qualifications or relevant and substantial research experience in an appropriate sector
  • Applicants must meet English requirements, or be able to do so before commencement

Please check the Higher Degree by Research minimum entry requirements.

Application Process

Applicants who require more information or are interested in this specific project should first contact the supervisor, Dr Son Tran.

Information and guidance on the application process can be found here.

To submit an application for this project, click here.