Prof. Sos Agaian, Distinguished Professor
Title: Quaternion Neural Networks: A Novel Framework for Robust Scene Perception in Self-Driving Cars
Abstract: Environment perception is crucial for safe autonomous driving, but it is often compromised by unpredictable impairments such as illumination, noise, and severe weather conditions. These impairments affect the performance of vision algorithms that handle tracking, detection, and segmentation tasks. This keynote talk presents Quaternion Neural Networks (QNNs) as a novel framework to overcome this challenge. QNNs operate on quaternion numbers, the four-dimensional generalizations of complex numbers, and have several benefits over real-valued neural networks. They process multidimensional inputs, capture internal dependencies such as between the color channels of an image pixel, model structural relations, reduce parameters, and avoid overfitting. The talk covers basics, such as quaternion convolutional and recent QNN transformers. It highlights their application in scene perception for self-driving cars under foggy, cloudy, and rainy environments. Attendees will see how QNNs surpass real-valued networks in weather removal and scene segmentation tasks.
Biography: Dr. Sos Agaian is a Distinguished Computer Science Professor at CUNY and a leading expert in bio-inspired artificial intelligence and cybersecurity systems. He holds a Ph.D. in Mathematics and a Dr. Engineering Science degree. He has significantly contributed to various domains, such as bioinspired AI, computational vision, cancer imaging, encryption-compression, and multimedia systems. His research aims to enable computers to perceive, learn, and understand the world as humans do. He has published over 750 articles, 10 books, 19 book chapters, and 56 patents/disclosures in prestigious journals and venues, including some related to COVID-19. He has mentored over 45 Ph.D. students and commercially licensed many of his intellectual properties. He is an Associate Editor for esteemed journals such as IEEE Transactions on Image Processing and IEEE Transactions on Cybernetics. He is a fellow of IS&T, SPIE, AAAS, IEEE, and AAIA and a Member of the Academia Europaea. He is also a Distinguished Lecturer in the SMC IEEE Society. He has given over 30 plenary/keynote speeches and more than 100 invited talks. He co-founded and served as an honorary/general co-chair and committee member for various international image processing and computer vision conferences. He received the Innovator of the Year Award in 2014 and the Tech Flash Titans-Top Researcher Award from the San Antonio Business Journal. He obtains research funding from prestigious agencies, such as AFRL, NSF, NIH, DARPA, the US Army, and the Department of Transportation.
Prof. Pong C Yuen
Department of Computer Science, Hong Kong Baptist University.
Title: Annotation-Efficient Deep Learning for Medical Image Segmentation
Abstract: Pixel-level annotation is an expensive process and has been one of the key challenges in developing deep learning based image segmentation. The problem is even more challenging for medical images because medical professional is required for annotation and normally abnormal regions in medical images are relatively small. In this talk, I will briefly give an overview of existing approaches in annotation-efficient deep learning for medical image segmentation. After that, I will report our approach by leveraging synthetic medical images. In specific, I will share three recent research works, namely weakly supervised learning using information extracted from radiology reports, synthetic-to-real test time training for source-free unsupervised domain adaption, and pseudo label guided image synthesis for semi-supervised learning. Finally, I will discuss the research directions towards annotation-free medical image segmentation.
Biography: Pong C Yuen received his Ph.D. degree in Electrical and Electronic Engineering in 1993 from The University of Hong Kong. He joined the Hong Kong Baptist University in 1993, and served as the Head of Department of Computer Science from 2011 – 2017. Currently, he is a Chair Professor in Computer Science and Associate Dean of Science Faculty, Hong Kong Baptist University.
Over the years, Dr. Yuen has spent his sabbatical and visited a number of universities/research institutes as a visiting professor/scholar, including The University of Sydney (Australia), The University of Maryland at College Park (USA), INRIA Rhone Alpes (France), ETH Zurich (Switzerland), and The University of Bologna (Italy). Dr. Yuen was the director of Croucher Advanced Study Institute (ASI) on biometric authentication in 2004 and the director of Croucher ASI on Biometric Security and Privacy in 2007. He has been serving as the Director of IAPR/IEEE Winter School on Biometrics since 2017.
Dr. Yuen has been actively involved in many international conferences and professional community in the capacity of general/program co-chair, including ICPR 2006, BTAS 2012, ISBA 2016, WIFS 2018, IJCB 2021. He served as Associate Editor of IEEE Transactions on Information Forensics and Security (2014 – 2018, 2021-2022), and received the Outstanding Editorial Board Service Award in 2018. Dr. Yuen has served as the Vice President (Technical Activities) of the IEEE Biometrics Council (2017-2019), and Associate Editor/Senior Editor of SPIE Journal of Electronic Imaging (2012-2019). Currently, Dr. Yuen is serving as the Senior Area Editor of IEEE Transactions on Information Forensics and Security, Editorial Board Member of Pattern Recognition, and Associate Editor of IEEE Transactions on Biometrics, Behaviour and Identity Science. He received the first-prize and second-prize Natural Science Awards from the Guangdong Province and the Ministry of Education, China, respectively. He is a Fellow of IAPR.
Dr. Yuen's current research interests include biometric security and privacy, video surveillance and medical informatics.
Prof. Sheli Sinha CHAUDHURI
Department of Electronics and Telecommunication Engineering, Jadavpur University, Kolkata, India
Title: Development of Land Cover Monitoring System using Simultaneous Multi-pixel Classification Strategy based Deep Neural Architecture
Abstract: Deep neural network based classification of satellite images have become an emerging area of research in recent years as it provides highly efficient ways to classify and interpret satellite imagery. Satellite images provide valuable information about the Earth's surface, such as land cover, vegetation, urban areas, and water bodies. By leveraging deep neural networks, it is possible to analyse and categorize these images with high accuracy and efficiency. Microwave remote sensing encompasses its benefits of providing cloud-free, all-weather images and images of day and night. Synthetic Aperture Radar (SAR) images own this characteristic which promoted the use of SAR and PolSAR images in land cover classification and various other applications for monitoring land use.
As part of a bigger project on land cover classification from polarimetric SAR images, I worked successfully on classification of vegetation areas growing diverse types of crops by designing a novel Convolutional Neural Network namely, Crop-Net, which facilitates simultaneous classification of multiple pixels located within an image patch taking into account the inter-relation between the scattering properties of neighbouring pixels. The efficiency of the designed network is validated using both publicly available data and real-field data and over-whelming performances are obtained in both cases.
Biography: Prof. Sheli Sinha CHAUDHIRI did her undergraduate, post graduate and doctoral studies from the department of Electronics and Telecommunication Engineering, Jadavpur University. She has been serving the same department as a faculty member since the last 23 years. She served as the Head of the Department during 2018-2020.She has been working in the field of Digital Image Processing since the last fifteen years. She did her doctoral studies on path planning of mobile robots based on the inputs of different sensors including a camera. Her research work includes developing various image processing algorithms, proceeding to the computer vision domain, mainly in the area of medical image analysis and restoration of weather degraded images. At present she is guiding her students in the automatic detection and classification of medical images, farm products, remote sensing images, waste images and tracking of objects from video sequences. Her research has resulted in a number of conference papers, journal papers, book chapters as well as copyrights on hazy night-time image benchmarking databases, a fuzzy controller for detection of sky regions in daytime outdoor images, and design of modified advanced Lee filter