Program Leader: Professor Massimo Piccardi
The primary objective of this program is to advance frontier research in the fields of computer vision, machine learning and multimedia, whilst maximising impact to industry.
Consisting of seven full-time academic staff and more than 30 PhD students, the group has an extensive track record of undertaking ARC projects, and conducting research for partner companies with funding from AusIndustry and Enterprise Connect.
Three dedicated computer vision labs are equipped with a range of advanced cameras, including depth cameras, high-frame rate, pan-tilt-zoom and long-range shortwave infrared.
Structured prediction in computer vision
Recognising multiple objects and activities in complex scenes is an open challenge in computer vision. Structured prediction tackles this challenge by exploiting the relationships amongst the objects and activities.
Text information extraction
Text appearing in outdoor scenes carries vital information for interpreting the scene’s contents and identifying objects and environments. Our research is leading technological advances in areas including license plate recognition, text sign recognition from natural scenes, and text information retrieval from images in web pages and emails. Our current focus is on deep features and deep learning architectures.
Multimedia for social media computing
Social multimedia analysis is in rapid expansion following the exponential growth of social networks and the associated data. Our research in this area is focused on multimedia content analysis, affective multimedia computing, multimedia recommender systems, multimedia indexing and retrieval, image retargeting and video adaptation, and social multimedia.
Perceptual interactive interfaces involving automated camera control
Nowadays cameras do not need to be static; for example, they can track a lecturer while they address an audience, or zoom in on passers-by in high security areas. This research area is developing advanced controls for perceptual, interactive interfaces that leverage pan-tilt-zoom cameras and computer vision.
Large-scale video surveillance, crowd analysis and gait recognition
Managing analysis of video surveillance content is a complex challenge. Research in this area is currently focused on analysis of videos from hundreds of cameras, movements and flows of large crowds, and recognition of individuals at a distance based on soft biometrics such as their gait.
Bayesian nonparametrics and general machine learning
Inference plays a fundamental role in computer vision. Our researchers are exploring current machine learning techniques relevant to computer vision and data mining, including Bayesian nonparametrics, structural SVM, variational inference, sub-modularity and others.