Overview
Background
Brian C. Lovell, born in Brisbane, Australia in 1960, received his BE in Electrical Engineering (Honours I) in 1982, BSc in Computer Science in 1983, and PhD in Signal Processing in 1991, all from the University of Queensland (UQ). Currently, he is the Project Leader of the Advanced Surveillance Group at UQ. Professor Lovell served as the President of the International Association of Pattern Recognition from 2008 to 2010, is a Senior Member of the IEEE, a Fellow of the IEAust, Fellow of the Asia-Pacific AI Association, and has been a voting member for Australia on the Governing Board of the International Association for Pattern Recognition since 1998.
He is an Honorary Professor at IIT Guwahati, India; an Associate Editor of the Pattern Recognition Journal; an Associate Editor-in-Chief of the Machine Learning Research Journal; a member of the IAPR TC4 on Biometrics; and a member of the Awards Committee and Education Committee of the IEEE Biometrics Council.
In addition, Professor Lovell has chaired and co-chaired numerous international conferences in the field of pattern recognition, including ICPR2008, ACPR2011, ICIP2013, ICPR2016, and ICPR2020. His Advanced Surveillance Group has collaborated with port, rail, and airport organizations, as well as several national and international agencies, to develop technology-based solutions for operational and security concerns.
His current research projects are in the fields of:
- Artificial Intelligence
- StyleGAN
- Stable Diffusion
- Deep Learning
- Biometrics
- Robust Face Recognition using Deep Learning
- Masked Face Recognition for COVID-19 Pandemic
- Adversarial Attacks on AI Systems
- Digital Pathology
- Neurofibroma Detection and Assessment
- Object Detection with Deep Learning
I am actively recruiting PhD students in Artificial Intelligence to work with my team. If you are interested and have a strong record from a good university, with a publication in a good conference such as CVPR, ICCV, ECCV, or MICCAI please send your CV to me. Full Scholarships (Tuition and Living) can be awarded within one month for truly exceptional candidates.
Availability
- Professor Brian Lovell is:
- Available for supervision
- Media expert
Fields of research
Qualifications
- Bachelor (Honours) of Engineering, The University of Queensland
- Bachelor of Science, The University of Queensland
- Doctor of Philosophy, The University of Queensland
Research interests
-
Face Recognition with Deep Learning
We develop new technologies to improve face recognition. Our group is first in the world to develop face recognition databases based entirely on synthetic faces. Other aspects of face recognition and affective computing (determining emotions from facial expressions) are current research themes within the group.
-
Object Detection Using Deep Learning
We are researching improved techniques to identify small objects with high precision
-
Synthetic Face and Image Generation
We were the first to investigate training face recognition systems on synthetic faces.
Research impacts
I have been pleased that my biometrics and other research has and is being been adopted commercially worldwide. My earlier face recognition systems have been installed by the University of San Francisco and Swinburne University among many other sites. More recently we have developed face recognition systems that are insensitive to the wearing of masks. These systems depend on our EDITH Ethical Face database of synthetic faces. To the best of our knowledge, we are the only group worldwide who can synthesise faces to order to train advanced ethical face recognition systems.
These systems have been adopted in the UK in 2020 by Facewatch Ltd and are currently being considered by the UK National Health Service and also Queensland Health to manage COVID 19 quarantine facilities and border control. In 2020-2021 we developed a touchless face mask fitting system for health workers to reduce the wastage of PPE and improve COVID19 management. This system is deployed on Queensland Health IT infrastructure in February 2021 and is planned to be made available nationally and internationally. The system has the potential to save millions of dollars in wasted PPE.
PRIZES, HONOURS AND AWARDS
Fellow of the IAPR, 2008 Multiple Best Paper prizes. Awarded Certificate of Recognition as most downloaded author at UQ by UQCybrary. Over 26,000 copies of my research papers were downloaded from the UQ EPrints archive in the 12 months ending May, 2005. APICTA Trophy for Best Research and Development, 2011, Face Recognition in a Crowd IFSEC Trophy 2011, Best CCTV Product of the Year (excluding cameras and lens), Face Recognition in a Crowd Technology Winner, ADS Security Innovation Award, 2021, Galahad facial detection and recognition software, awarded by the UK Home Office at the Security and Policing Show on March 9, 2021.
Works
Search Professor Brian Lovell’s works on UQ eSpace
2024
Journal Article
Domain-aware triplet loss in domain generalization
Guo, Kaiyu and Lovell, Brian C. (2024). Domain-aware triplet loss in domain generalization. Computer Vision and Image Understanding, 243 103979, 103979. doi: 10.1016/j.cviu.2024.103979
2023
Conference Publication
Uncertainty-guided cross-modality semi-supervised learning for MRI segmentation
Li, Chaoyi and Lovell, Brian C. (2023). Uncertainty-guided cross-modality semi-supervised learning for MRI segmentation. 2023 International Conference on Digital Image Computing: Techniques and Applications (DICTA), Port Macquarie, NSW, Australia, 28 November - 1 December 2023. Piscataway, NJ, United States: IEEE. doi: 10.1109/dicta60407.2023.00018
2023
Journal Article
Antisunward streaming westward Sub‐Auroral Ion Drifts (SAID) developed in the postmidnight (1‐4) magnetic local time sector during 2013
Horvath, Ildiko and Lovell, Brian C. (2023). Antisunward streaming westward Sub‐Auroral Ion Drifts (SAID) developed in the postmidnight (1‐4) magnetic local time sector during 2013. Journal of Geophysical Research: Space Physics, 128 (9) e2023JA031808. doi: 10.1029/2023ja031677
2023
Conference Publication
Knowing the unknown: open-set bacteria classification in gram stain microscopic images
Alhammad, Sarah and Lovell, Brian C. (2023). Knowing the unknown: open-set bacteria classification in gram stain microscopic images. 2023 International Joint Conference on Neural Networks (IJCNN), Gold Coast, QLD, Australia, 18-23 June 2023. Piscataway, NJ, United States: Institute of Electrical and Electronics Engineers. doi: 10.1109/ijcnn54540.2023.10191073
2023
Journal Article
Sub‐Auroral Ion Drifts (SAID) Developed Over the Northern Winter Hemisphere at Dawn During 2016–2017
Horvath, Ildiko and Lovell, Brian C. (2023). Sub‐Auroral Ion Drifts (SAID) Developed Over the Northern Winter Hemisphere at Dawn During 2016–2017. Journal of Geophysical Research: Space Physics, 128 (4) e2022JA031228, 1-18. doi: 10.1029/2022ja031228
2023
Journal Article
DIODE: dilatable incremental object detection
Peng, Can, Zhao, Kun, Maksoud, Sam, Wang, Tianren and Lovell, Brian C. (2023). DIODE: dilatable incremental object detection. Pattern Recognition, 136 109244, 109244. doi: 10.1016/j.patcog.2022.109244
2023
Journal Article
Subauroral flows and associated magnetospheric and ionospheric phenomena developed during 7‐8 September 2017
Horvath, Ildiko and Lovell, Brian C. (2023). Subauroral flows and associated magnetospheric and ionospheric phenomena developed during 7‐8 September 2017. Journal of Geophysical Research: Space Physics, 128 (3) e2022JA030966, 1-20. doi: 10.1029/2022ja030966
2023
Journal Article
Abnormal Sub‐Auroral Ion Drifts (ASAID) developed in various inner‐magnetosphere configurations at geomagnetically quiet times
Horvath, Ildiko and Lovell, Brian C. (2023). Abnormal Sub‐Auroral Ion Drifts (ASAID) developed in various inner‐magnetosphere configurations at geomagnetically quiet times. Journal of Geophysical Research: Space Physics, 128 (1) ARTN e2022JA031004, 1-19. doi: 10.1029/2022ja031004
2023
Conference Publication
Dynamic curriculum learning via in-domain uncertainty for medical image classification
Li, Chaoyi, Li, Meng, Peng, Can and Lovell, Brian C. (2023). Dynamic curriculum learning via in-domain uncertainty for medical image classification. MICCAI 2023: 26th International Conference, Vancouver, BC, Canada, 8-12 October 2023. Cham, Switzerland: Springer Nature Switzerland. doi: 10.1007/978-3-031-43904-9_72
2023
Conference Publication
End to end generative meta curriculum learning for medical data augmentation
Li, Meng, Li, Chaoyi, Peng, Can, Liu, Liangchen and Lovell, Brian (2023). End to end generative meta curriculum learning for medical data augmentation. 30th IEEE International Conference on Image Processing (ICIP), Kuala Lumpur, Malaysia, 8-11 October 2023. Piscataway, NJ, United States: Institute of Electrical and Electronics Engineers. doi: 10.1109/ICIP49359.2023.10222093
2022
Journal Article
Duskside sub‐auroral polarization streams (SAPS) and dawnside subauroral flows during the magnetically quiet 24 November and moderately active 25–27 November 2008
Horvath, Ildiko and Lovell, Brian C. (2022). Duskside sub‐auroral polarization streams (SAPS) and dawnside subauroral flows during the magnetically quiet 24 November and moderately active 25–27 November 2008. Journal of Geophysical Research: Space Physics, 127 (12) ARTN e2022JA030609. doi: 10.1029/2022ja030609
2022
Journal Article
Newly formed dawnside, duskside, and nightside subauroral flows developed during magnetically active times
Horvath, Ildiko and Lovell, Brian C. (2022). Newly formed dawnside, duskside, and nightside subauroral flows developed during magnetically active times. Journal of Geophysical Research: Space Physics, 127 (10) e2021JA030215, 1-21. doi: 10.1029/2021ja030215
2022
Conference Publication
Efficient cell labelling for gram stain WSIs
Alhammad, Sarah, Zhang, Teng, Zhao, Kun, Hobson, Peter, Jennings, Anthony and Lovell, Brian C. (2022). Efficient cell labelling for gram stain WSIs. 26th International Conference on Pattern Recognition (ICPR), Montreal, QC, Canada, 21-25 August 2022. Piscataway, NJ, United States: Institute of Electrical and Electronics Engineers. doi: 10.1109/icpr56361.2022.9956490
2022
Conference Publication
MedViTGAN: end-to-end conditional GAN for histopathology image augmentation with vision transformers
Li, Meng, Li, Chaoyi, Hobson, Peter, Jennings, Tony and Lovell, Brian C. (2022). MedViTGAN: end-to-end conditional GAN for histopathology image augmentation with vision transformers. 26th International Conference on Pattern Recognition / 8th International Workshop on Image Mining - Theory and Applications (IMTA), Montreal, Canada, 21-25 August 2022. New York, NY USA: Institute of Electrical and Electronics Engineers. doi: 10.1109/icpr56361.2022.9956431
2022
Conference Publication
Few-Shot Class-Incremental Learning from an Open-Set Perspective
Peng, Can, Zhao, Kun, Wang, Tianren, Li, Meng and Lovell, Brian C. (2022). Few-Shot Class-Incremental Learning from an Open-Set Perspective. 17th European Conference on Computer Vision - ECCV 2022, Tel Aviv, Israel, 23–27 October 2022. Cham, Switzlerland: Springer Nature. doi: 10.1007/978-3-031-19806-9_22
2021
Conference Publication
Efficient DNN-Based Classification of Whole Slide Gram Stain Images for Microbiology
Alhammad, Sarah, Zhao, Kun, Jennings, Anthony, Hobson, Peter, Smith, Daniel F., Baker, Brett, Staweno, Justin and Lovell, Brian C. (2021). Efficient DNN-Based Classification of Whole Slide Gram Stain Images for Microbiology. 2021 Digital Image Computing: Techniques and Applications (DICTA), Gold Coast, QLD Australia, 29 November - 1 December 2021. Piscataway, NJ United States: IEEE. doi: 10.1109/dicta52665.2021.9647415
2021
Journal Article
SID: Incremental learning for anchor-free object detection via Selective and Inter-related Distillation
Peng, Can, Zhao, Kun, Maksoud, Sam, Li, Meng and Lovell, Brian C. (2021). SID: Incremental learning for anchor-free object detection via Selective and Inter-related Distillation. Computer Vision and Image Understanding, 210 103229, 103229. doi: 10.1016/j.cviu.2021.103229
2021
Journal Article
Investigating the coupled Magnetosphere‐Ionosphere‐Thermosphere (M‐I‐T) system’s responses to the 20 November 2003 superstorm
Horvath, Ildiko and Lovell, Brian C. (2021). Investigating the coupled Magnetosphere‐Ionosphere‐Thermosphere (M‐I‐T) system’s responses to the 20 November 2003 superstorm. Journal of Geophysical Research: Space Physics, 126 (9) e2021JA029215. doi: 10.1029/2021ja029215
2021
Journal Article
Subauroral flow channel structures and auroral undulations triggered by Kelvin‐Helmholtz waves
Horvath, Ildiko and Lovell, Brian C. (2021). Subauroral flow channel structures and auroral undulations triggered by Kelvin‐Helmholtz waves. Journal of Geophysical Research: Space Physics, 126 (6) e2021JA029144. doi: 10.1029/2021ja029144
2021
Conference Publication
Minimizing labeling cost for nuclei instance segmentation and classification with cross-domain images and weak labels
Yang, Siqi, Zhang, Jun, Huang, Junzhou, Lovell, Brian C. and Han, Xiao (2021). Minimizing labeling cost for nuclei instance segmentation and classification with cross-domain images and weak labels. 35th AAAI Conference on Artificial Intelligence, AAAI 2021, Virtual, 2-9 February 2021. Menlo Park, CA, United States: Association for the Advancement of Artificial Intelligence. doi: 10.1609/aaai.v35i1.16150
Funding
Current funding
Supervision
Availability
- Professor Brian Lovell is:
- Available for supervision
Before you email them, read our advice on how to contact a supervisor.
Available projects
-
Detecting and Classifying Neurofibromas using Deep Learning
Neurofibromatosis type 1 (NF1) is one of the most common single-gene inherited disorders globally, with an incidence of 1/2500 individuals. While several phenotypes are associated with the disorder, the most common manifestation is cutaneous neurofibroma. The majority of adults develop these distressing cutaneous tumours (cNF), which increase in severity with age. Adult patients report cosmetic disfigurement due to cNF as the greatest burden of living with NF1. There is no way to predict tumour severity which can range from <100 to thousands. Youth and families experience reduced quality of life due to concerns about this uncertain future. We don’t yet understand why this condition is so variable or have any effective medical treatments. In the proposed research, we will assemble a consortium of internationally recognised experts in NF1 with access and capacity to recruit and phenotype patients to drive the largest genome-wide association and epigenome-wide association studies of the modifier gene networks driving the cutaneous phenotypic variance in NF1. We will then use individualised pharmacological annotation of these networks to identify precision treatment options to mitigate the most distressing and life quality damaging aspects of this devastating illness.
-
Classifying Gram Stain Images Using Transformers and Deep Learning
Microscopic diagnosis of Gram stain smears is one of the most time and labor intensive tasks in the clinical setting. With the recent development of automated digital pathology scanners, it is now possible to economically obtain high-resolution Gram stain whole slide images for routine diagnosis. This finally opens the doorway to automated identification of bacteria types from digital images in a clinical setting. However, Gram stain whole slide images comprise billions of pixels and suffer from high morphological heterogeneity as well as from many different types of artifacts. Identifying multiple types of tiny bacteria with various densities from an extremely large whole slide image is incredibly challenging. To this end, we propose an end-to-end framework with a novel loss function that tackles the patch aggregation while considering the correlation of different labels in this multi-label scenario. Our framework first effectively integrates the relations among multiple patch features, and then utilizes a class aggregator to generate a robust slide-level feature representation under multi-label setting. Furthermore, we propose a novel loss function integrating two regularization terms: 1) a negative regulator that reduces the confusion between bacteria and negative samples without any bacteria, and 2) an adversarial loss that alleviates the impact of background specification among various tissue samples. We show that the proposed method achieves superior performance compared to several state-of-the-art methods.
-
Incremental Learning for AI
Incremental learning requires a model to continually learn new tasks from streaming data. However, traditional fine-tuning of a well-trained deep neural network on a new task will dramatically degrade performance on the old task — a problem known as catastrophic forgetting. We address this issue in the context of anchor-free object detection, which is a new trend in computer vision as it is simple, fast, and flexible. Simply adapting current incremental learning strategies fails on these anchor-free detectors due to lack of consideration of their specific model structures. To deal with the challenges of incremental learning on anchor-free object detectors, we propose a novel incremental learning paradigm called Selective and Inter-related Distillation (SID). In addition, a novel evaluation metric is proposed to better assess the performance of detectors under incremental learning conditions. By selective distilling at the proper locations and further transferring additional instance relation knowledge, our method demonstrates significant advantages on the benchmark datasets PASCAL VOC and COCO.
-
Text to Face Synthesis using Stable Diffusion for Biometrics Research
Text-to-Face (TTF) synthesis is a challenging task with great potential for diverse computer vision applications. Compared to Text-to-Image (TTI) synthesis tasks, the textual description of faces can be much more complicated and detailed due to the variety of facial attributes and the parsing of high dimensional abstract natural language. We propose a text-to-face model that should not only produce images in high resolution (10241024) and text-to-image consistency, but also output multiple faces to cover a wide range of unspecified facial features in a natural way. By fine-tuning the multi-label classifier and image encoder, our model obtains the vectors and image embeddings which are used to manipulate the noise vector sampled from the normal distribution. Afterwards, the manipulated noise vector is fed into a pre-trained high-resolution image generator to produce a set of faces with desired facial attributes. We refer to our model as TTF-HD. Experimental results show that TTF-HD generates high-quality faces and achieves state-of-the-art performance.
Supervision history
Current supervision
-
Doctor Philosophy
Medical Image Segmentation with Limited Annotated Data
Principal Advisor
Other advisors: Associate Professor Marcus Gallagher
-
Doctor Philosophy
The role of duality in machine learning and computer vision.
Principal Advisor
Other advisors: Dr Mahsa Baktashmotlagh
-
Doctor Philosophy
Enhancing Building Fire Safety by Utilising Machine Learning Techniques
Associate Advisor
Other advisors: Dr Xin Yu
-
Doctor Philosophy
Modelling cloud movement to generate short term solar irradiance predictions and subsequent expected PV power production
Associate Advisor
Other advisors: Professor Eve McDonald-Madden, Dr Hui Ma
-
Doctor Philosophy
Generating data-driven continuous optimization problems for benchmarking
Associate Advisor
Other advisors: Associate Professor Marcus Gallagher
Completed supervision
-
2024
Doctor Philosophy
Data Augmentation through Image Synthesis and Editing in Multi-domains via Disentangled Latent Space
Principal Advisor
-
2024
Doctor Philosophy
Modelling Efficient and Robust Solutions for Microbiology Image Analysis Using Deep Learning
Principal Advisor
-
2023
Doctor Philosophy
Whole Slide Image Processing in the Immunofluorescence Domain
Principal Advisor
-
2023
Doctor Philosophy
Modeling Data Scarcity Solutions with Deep Learning for Histopathology Image Analysis
Principal Advisor
-
-
2020
Doctor Philosophy
On the Robustness of Object and Face Detection: False Positives, Attacks and Adaptability
Principal Advisor
-
2020
Master Philosophy
Accurate Dense Depth From Light Field Technology For Object Segmentation And 3D Computer Vision
Principal Advisor
-
2017
Doctor Philosophy
Exploring The Frontier of Smart Video Surveillance: Novel Domains and Fine-Grain Event Understanding
Principal Advisor
-
-
2016
Doctor Philosophy
Fast and Accurate Image and Video Analysis on Riemannian Manifolds
Principal Advisor
-
2016
Doctor Philosophy
Techniques for Binocular Markerless Visual Tracking of 3D Articulated Bodies
Principal Advisor
-
2016
Doctor Philosophy
Action Analysis and Video Summarisation to Efficiently Manage and Interpret Video Data
Principal Advisor
-
2016
Doctor Philosophy
Countering Digital Replay Attacks for Face Verification on Consumer Smart Devices using Structured Illumination
Principal Advisor
Other advisors: Emeritus Professor Paul Bailes
-
-
-
-
2014
Doctor Philosophy
Towards Robust Representation and Classification of Low-resolution Image and Video Data
Principal Advisor
-
2014
Doctor Philosophy
Video Analysis Based on Learning on Special Manifolds for Visual Recognition
Principal Advisor
-
2013
Doctor Philosophy
Improving Representation and Classification of Image and Video Data for Surveillance Applications
Principal Advisor
-
2012
Doctor Philosophy
Contextual Spatial Analysis and Processing for Visual Surveillance Applications
Principal Advisor
-
2012
Doctor Philosophy
Towards Robust Identity Inference Under Surveillance Environments: From Still Images to Video Sequences
Principal Advisor
-
2011
Doctor Philosophy
Real-Time Face Detection on High Resolution Smart Camera System
Principal Advisor
-
-
2008
Master Philosophy
Robust Person and Vehicle Tracking for Intelligent Visual Surveillance
Principal Advisor
-
2008
Doctor Philosophy
Efficient and invariant regularisation with application to computer graphics
Principal Advisor
-
2007
Doctor Philosophy
ROBUST FACE RECOGNITION FOR VIDEO SURVEILLANCE USING A SINGLE GALLERY IMAGE
Principal Advisor
-
2006
Doctor Philosophy
EFFICIENT METHODS FOR 3D RECONSTRUCTION FROM MULTIPLE IMAGES
Principal Advisor
-
2006
Doctor Philosophy
THE MULTIPLE-VIEW GEOMETRY OF IMPLICIT CURVES AND SURFACES
Principal Advisor
-
2005
Doctor Philosophy
Training Hidden Markov Models for Spatio-Temporal Pattern Recognition
Principal Advisor
-
2005
Doctor Philosophy
ROBUST DISCRIMINATIVE PRINCIPAL COMPONENT ANALYSIS FOR FACE RECOGNITION
Principal Advisor
-
2005
Doctor Philosophy
GLOBALLY MINIMAL CONTOURS AND SURFACES FOR IMAGE SEGMENTATION
Principal Advisor
-
2004
Master Engineering Sc
OPTICAL FLOW FROM COLOUR IMAGE SEQUENCES
Principal Advisor
-
2004
Doctor Philosophy
HAND GESTURE RECOGNITION BY HIDDEN MARKOV MODELS
Principal Advisor
-
2003
Master Philosophy
Three-dimensional reconstruction of highly complex and mobile surrounds from stereo images and image sequences
Principal Advisor
-
-
2008
Doctor Philosophy
A SEMANTIC FRAMEWORK FOR THE MANAGEMENT, ANALYSIS AND ASSIMILATION OF MIXED-MEDIA SCIENTIFIC DATA
Associate Advisor
-
2006
Doctor Philosophy
AUTOMATIC BANK CHECK PROCESSSING AND AUTHENTICATION USING SIGNATURE VERIFICATION
Associate Advisor
-
2004
Master Philosophy
SECURE LOCATION SERVICES - VULNERABILITY ANALYSIS AND PROVISION OF SECURITY IN LOCATION SYSTEMS
Associate Advisor
-
2004
Master Philosophy
COLLABORATIVE VIDEO INDEXING, ANNOTATION AND DISCUSSION OVER HIGH-BANDWIDTH NETWORKS
Associate Advisor
Media
Enquiries
Contact Professor Brian Lovell directly for media enquiries about:
- Artificial Intelligence
- Biometrics
- Border control
- Computer modelling
- Computer vision
- Deep Learning
- Face Recognition
- Face-recognition technology
- Identification technology
- Image processing
- Information technology
- National security surveillance
- Networks - neural
- Neural networks - artificial
- Pattern Recognition
- Pattern recognition - digital imaging
- Signal Processing
- Wearable Technologies
Need help?
For help with finding experts, story ideas and media enquiries, contact our Media team: