Current research focus (March 2024)
Medical Report Generation (Or Radiology Report Generation)
The objective of medical report generation is to create computational algorithms capable of generating textual reports that articulate the findings and impressions in the images taken in modalities such as chest X-rays, CT scans, MRIs, or ultrasounds. This endeavor holds the promise of substantially alleviating the current workload burden on radiologists, enhancing their operational efficiency, mitigating examination errors, and ultimately facilitating the delivery of superior and more sustainable healthcare services to society. The primary aim of this research is to advance the development of sophisticated approaches, methodologies, and algorithms for the representation of visual and textual information, and the seamless translation of the former into the latter. Deep learning techniques, alongside large language and vision models, as well as knowledge representation and reasoning methodologies, are poised to play pivotal roles in this pursuit.
Fine-grained Vision and Language Understanding and Interaction
The last decade has witnessed the significant progress on understanding vision, language and their interactions, with the advent of large models being a good evidence. Nevertheless, the current models are still far from being satisfactory in terms of fine-grained understanding, that is, capturing the subtle, detailed contents of image and text and associating them in a precise manner. Having the capabilities is highly desirable in many tasks involving fine-grained image and text understanding and the generation between image and text, especially when the background applications are medicine- or healthcare-related. This research aims to develop advanced approaches, methodologies, and techniques to understand image and text in a more nuanced way and better align visual and textual information in order to achieve comprehensive and accurate understanding.
Continual Learning with the Application to Robotics
This project endeavours to pioneer novel machine learning techniques aimed at enhancing machines' ability to leverage past experiences to tackle new tasks with limited data. Its primary objective is to mitigate the undesirable reliance of current machine learning methodologies on labeled data while substantially enhancing their performance, particularly in applications related to robotics. Anticipated outcomes of the project include the generation of new theoretical insights into continual learning, alongside the development of innovative algorithms poised to underpin the creation of the next generation of computer vision, machine learning, and robotics systems tailored to operate within open and dynamic environments. The envisioned impact of these advancements extends to benefiting science, society, and the economy through the application of these advanced intelligent systems. Deep learning, meta-learning, reinforcement learning, and the utilisation of large-scale models are expected to play pivotal roles in driving the success of this research endeavour.
Generic Image recognition
Image and object recognition has recently made significance progress with the advent of deep learning. The performance and efficiency of visual recognition have been improved to an unprecedentedly high level. Our recent research in this regard is focused on i) symmetric positive-definite (SPD) matrix-based visual representation; ii) unsupervised domain adaptation for image classification; and ii) new models and algorithms for fine-grained image recognition.
Content-based Image Retrieval
Content-based image retrieval has recently witnessed its fast development due to the pervasive use of mobile platform and the explosion of the volumes of images over the Internet. Our recent research in this regard is focused on i) query-adaptive image retrieval; ii) image retrieval via unsupervised deep learning; and iii) retrieval on archival photographic collections, by collaborating with national and regional organisations.
Grants and Awards
2023 UOW Research Partnership Grant Scheme (AUD19,728, 2024), Using AI to Enhance Future Chemical Manufacturing, Chris Hyland, Sinead Keaveney, Lei Wang, Chris Richardson and Mark Waller
2022 ARC Discovery Project, Australian Research Council (AUD450,000, 2022—2024), Making Meta-learning Generalised, Lei Wang, Peng Wang, and Lingqiao Liu
2020 UOW-BUAA Collaborative Grant (AUD5,000), Developing Extremely Efficient Deep Neural Networks for Long-tail Image Recognition, Lei Wang and Peng Wang
2020 NHMRC Ideas Grant, NHMRC (AUD823,476.40, 2020—2022), The algorithm will see you now: ethical, legal and social implications of adopting machine learning systems for diagnosis and screening, Role: One of Chief Investigators
2020 ARC Discovery Project, Australian Research Council (AUD445,000, 2020—2022), Learning kernel-based high-order visual representation for image retrieval, Role: Sole Chief Investigator
2019 Global Challenges Project Funding University of Wollongong (AUD49,239), The ethical, legal and social implications (ELSI) of using artificial intelligence (AI) in health and social care, Role: One of Chief Investigators
2018 Advantage SME Collaboration Voucher, NSW Department of Industry (AUD60,000), Smarter building sensor, control and data analysis systems. Role: One of Chief Investigators
2018 Faculty EIS Strategic Investments Grant for DP, University of Wollongong (AUD10,000), Cost-effective pathogen-host protein-protein interactions, Lei Wang and Jun Shen
2017 Community Engagement Grant, University of Wollongong (AUD8,000), Retrieving lost community stories: linking regional archival photo collections using advanced visual technologies, Role: Lead Chief Investigator
2019-2016 National Computational Merit Allocation Scheme, Exploring National Treasure: Retrieval of Large Collection of Archival Photos, Role: Lead Chief Investigator
2014-15 Commercial Research Project with an IT Company in US (AUD182,000), Clothing Image Retrieval Research, Role: Lead Chief Investigator
2013 Near Miss Grant for ARC Discovery Project, University of Wollongong (AUD10,000), Role: Lead Chief investigator
2012 Near Miss Grant for ARC Discovery Project, University of Wollongong (AUD12,000), Role: Lead Chief investigator
2012 URC Small Grant, University of Wollongong University Research Committee (AUD13,000), When Vision Meets Fashion: Addressing The Problem of Fine-grained Image Retrieval, Lei Wang and Markus Hagenbuchner
2011 Research Develop Scheme Grant, University of Wollongong Faculty of Informatics (AUD3,500), Role: Sole investigator
2011 Research Infrastructure Block Grant, University of Wollongong (AUD70,000), 3D Multimedia Research Infrastructure, Role: Coordinator / Primary User
2009 Early Career Researcher Award, Australian Academy of Science Australian Research Council
2009 ARC Linkage Project, Australian Research Council (AUD240,000 from ARC, 2009—2012), Generic Content-based Picture Retrieval with Its Application to Archival Photographic Collections, Lei Wang, Richard Hartley, and Hongdong Li
2007 ARC Discovery Project, Australian Research Council (AUD255,000, 2007—2009 ), Computer Vision Optimization Problem Using Machine Learning, Richard Hartley and Lei Wang (Australian Postdoctoral Fellowship)
HDR Principal Supervisor (topics may change)
(On-going) Sangrasi Din Muhammad, PhD student, UOW, Large-scale Facial Image Clustering for Archival Photo Collection
(On-going) Ian Comor, MPhil student, UOW, Integrating Textual and Visual Information for Archival Photo Retrieval
----------------------------------------------------------------------------------------------------------
(Completed) Saimunur Rahman, PhD student, UOW, Fine-grained Image Recognition and Its Applications
(Completed) Yu Ding, PhD student, UOW, Deep Transfer Learning and Its Applications
(Completed) Zhongyan Zhang, PhD student, UOW, Sample-adaptive Learning and Its Applications
(Completed) Biting Yu, PhD student, UOW (Jointly supervised with Dr. Luping Zhou), Deep Learning Techniques for Medical Image Analysis
(Completed) Zhexuan Zhou, MPhil student, UOW, Microatoll Detection for Great Barrier Reef with Deep Neural Networks
(Completed) Bela Chakraborty, MPhil student, UOW, Unsupervised Learning of Archival Photo Collection for Retrieval
(Completed) Yan Zhao, PhD student, UOW, Deep Learning: Theories and Applications
(Complete) Melih Engin, MPhil student, UOW, Deep SPD Representation for Visual Recognition
(Completed) Zhimin Gao, PhD student, UOW, Developing Advanced Deep Learning Models for Visual Recognition
(Completed) Jianjia Zhang, PhD student, UOW, Medical Image Analysis with Advanced Visual Recognition Models
(Completed) Xinwang Liu, PhD student, ANU, Machine Learning Algorithm Research on Multiple Kernel Learning
(Completed) Lingqiao Liu, PhD student, ANU, Recognizing Human Actions in Video Sequences
(Completed) Jixi Yang, Honours student, UOW, Content-based Image Retrieval on Archival Photo Collections
(Completed) Yifan Lu, PhD student, ANU, Inferring Human Pose and Motion from Images and Videos
(Completed, HD) Aisha Khan, Master student, ANU, Detecting Unusual Events in Indoor Smart Surveillance
(Completed, HD) Wenbo Zhao, Honours student, ANU, A Demonstration System of Content-based Image Retrieval
(Completed) Chuan Hu, Masters student, ANU, Improving SIFT features to Handle Significant Scale Difference for Object Retrieval
(Completed) Yuhang Zhang, MPhil student, ANU, Local Invariant Feature based Object Retrieval in a Supermarket
(Completed, HD) Chuan Hu, Honours student, ANU, Evaluation of Local Invariant Features for Generic Content-based Image Retrieval
(Completed) Ran Huo, Honours student, ANU, Evaluation of Local Invariant Features for Generic Image Retrieval with Relevance Feedback