Dr Yang Song

Key areas: biomedical image analysis, general computer vision, deep learning, machine learning, interdisciplinary computer vision applications, multimedia data analysis

Actively recruiting PhD students. Interested candidates are strongly encouraged to apply.

Example topics are listed below. You are welcome to propose alternative research topics in the computer vision field.

Morphology analysis in microscopy images

Various types of microscopy images are widely used in biological research to aid our understanding of human biology. Cellular and molecular morphologies give lots of information about the underlying biological processes. The ability to identify and describe the morphological information quantitatively, objectively and efficiently is critical. In this PhD project, we will investigate various computer vision, machine learning (especially deep learning) and statistical analysis methodologies to develop automated morphology analysis methods for microscopy images.

Cell segmentation and event detection in microscopy images

One type of morphology analysis approach focuses on the cell-level information extraction from microscopy images. Normal cells, cancer cells, and special events (such as mitosis) are the important objects of interest. Detection and segmentation of these cells are challenging with the large variety of tissues and inconsistent imaging conditions. In this project, we will investigate various segmentation algorithms with a particular focus on machine/deep learning.

Lesion detection and segmentation in MRI

Detection and segmentation of lesions in MRI are routinely performed in radiology centres for patient diagnosis and treatment planning. This is a time-consuming process and prone to inter-observer variability. Computerized methods have been developed over the years but there is still much scope of improvement. In this PhD project, we will investigate various types of lesions, such as brain tumours, with advanced deep learning algorithms and other techniques such as deformable models.

Object detection and fine-grained recognition

This is an emerging topic in general computer vision. The objective is to detect the object (e.g. a cat) and classify the object at a fine-grained level (e.g. a siamese cat), and such systems would have a large variety of practical applications in real life scenarios. Recently lots of research progress have been reported in this area. In this PhD project, we will investigate the advanced methodologies, which mainly involve various types of deep learning models, with a particular focus to address the challenges associated with insufficient sample sizes. 

Weakly supervised object localization

With limited image-level annotations, a weakly-supervised approach is essential to localize the objects of interest in images. Such a method helps to reduce the amount of detailed ground truth annotation, which is traditionally necessary for fully-supervised machine or deep learning methods. In this PhD project, we will investigate the advanced methodologies, which mainly involve various types of deep learning models, with a wide variety of application areas in both general and biomedical image domains. 

Vision-based human behaviour analysis

Human behaviour analysis covers a wide range of topics and is an essential step for many autonomous platforms such as self-driving cars and social robots. Different approaches have been proposed over the recent years with novel development of various deep learning models. In this PhD project, we will work on a particular type of human behaviour analysis, investigating advanced methodologies, with a particular focus on integrating social contexts and knowledge representation into the deep learning models.