Invited Speakers

Prof. Wei Fang
Nanjing University of Information Science & Technology, China

Wei Fang is currently a Professor of Computer Science at the NanJing University of Information Science &Technology in China. He was a Visiting scholar of University of Florida, USA. Dr. Fang is a Senior Member of China Computer Society, a member of IEEE, a member of ACM, a member of Jiangsu Artificial Intelligence Society, and is also a Science and Technology Consulting Expert of Jiangsu Province, a member of Artificial Intelligence Professional Committee of Shanghai Meteorological Society. He is the founding Director of Institute of Artificial Intelligence and Meteorological Big Data. He organized several conferences. He is a guest editor and editorial board member for several SCI journals.

Speech Title: Research on the Application of Artificial Intelligence in Meteorology
Artificial intelligence (AI) has become widely used in a variety of fields, including disease prediction, environmental monitoring, and pollutant prediction. In recent years, there has also been an increase in the volume of research into the application of AI to Meteorology .This speech aims to explore the latest trends in the application of AI in Meteorology.

The Artificial Intelligence (AI) is playing a more and more essential role in the industrial revolution and we are seeking a lot of evolution in various machine learning methodologies. Forecast of meteorological disasters is an important and challenging worldwide problem. Various techniques have been used to solve it, but the accuracy of them is not high due to the highly nonlinear, random, and complex nature of precipitation. In recent years, with the rapid development of artificial intelligence technology, it has gradually penetrated into all aspects of people's lives, and the meteorological field are no exception. This speech will share our experience and research results in all aspects of the application of meteorology based on artificial intelligence.

Prof. Zhenbo Li
China Agricultural University, China

Zhenbo Li, professor, doctoral supervisor, vice dean of College of Information and Electrical Engineering, China Agricultural University. His main research direction is computer vision and smart agriculture. He has presided over 8 national research projects such as National Natural Science Foundation of China and National Key Research and Development Program, 8 provincial and ministerial level research projects. He has published more than 90 high-level papers in academic journals and top academic conferences(T-IP, ICCV, ECCV etc.) , obtained more than 40 national invention patents, more than 60 software copyrights, and won 4 provincial and ministerial science and technology awards.

Speech Title: Data-driven Image Restoration and Underwater Image Enhancement
The data-driven artificial intelligence model has become a new research paradigm. This presentation presents a summary of the latest research on the image restoration (IR) and underwater image enhancement (UIE) tasks. I will introduce some of our related research. Most existing image-restoration models employed static CNN-based models, where the fixed learned filters cannot fit the diverse degradation well. To address this, we propose a novel dynamic image restoration contrastive network (DRCNet). Image restoration aims to recover images from spatially-varying degradation. In contrast, UIE aims to reconstruct clear images from the diverse degradation images caused by light scattering and attenuation in underwater scenarios. However, existing UIE methods always face the dilemma of the trade-off between performance and efficiency. Therefore, we design a novel invertible UIE network, InvUIE, that can optimally balance these competing goals. Moreover, due to the degraded underwater images have hampered the existing computer vision applications. We designed a novel tied bilateral learning network is proposed for UIE, dubbed TBAIE, which improves the degraded underwater images to meet the requirements of various computer vision applications, such as aquaculture etc..

Prof. Xin Xu
Wuhan University of Science and Technology, China

Xin Xu received the B.S. and Ph.D. degrees in computer science and engineering from Shanghai Jiao Tong University, Shanghai, China, in 2004 and 2012 respectively. He is currently a Full Professor with the School of Computer Science and Technology, Wuhan University of Science and Technology, Wuhan, China. His research deals with image processing, computer vision, and deep learning. More specifically, his research areas focus on building a hierarchical person re-identification architecture including detection and recognition for nighttime surveillance scenarios. His publication was selected as the cover paper of the journal International Journal of Intelligent Systems in 2022. He was shortlist in the Best Paper Finalist of the IEEE International Conference on Multimedia and Expo (ICME) 2021.

Speech Title: Non-Convex Machine Iearning based Large-Scale Interactive Image Annotation
The rapid development of artificial intelligence in recent years can be seen as a triumph of data intelligence, where machine learning plays a critical role. Most of current machine learning techniques are typically based on solving convex optimization problems. Recently, many real-world applications that require solving non-convex optimization problems are continuously emerging, which can be much more challenging. Typical applications include interactive data annotation, deep network compression, and etc. Traditionally, researchers have attempted to approximate non-convex optimization problems as convex problems in order to apply standard optimization techniques. However, these approximations can be quite crude and may not capture the full complexity of the original problem. To effectively promote the potential of machine learning to solve practical application, it is urgent to directly handle non-convex problems from the source.

Prof. Zhenlong Du
Nanjing Tech University, China

Zhenlong Du is currently a Professor of Computer Science at NanJing Tech University in China. He was a Visiting scholar of SUNNY at Buffalo, USA. Dr. Du is a Senior Member of China Computer Society, a member of IEEE, a member of ACM, and a member of Jiangsu Computer Society. Dr. Du is honored the 3rd prize of Jiangsu Science and Technology Award in 2020, he is honored the Jiangsu Six Talent Peaks. Recently, Dr. Du has hosted National Natural Science Foundation of China, Higher Education Doctoral Program of the Ministry of Education, etc. Meanwhile, Dr. Du is hosting several industrial research projects.

Speech Title: Fusion of Edge and Cloud in Manufacture Updating and Digitalized Promotion
The report presents the fusion practice between AI model and analytic modeling in manufacture updating and digitalized promotion. It analyzes the coordination of edge along with cloud based on their respective merits. The presentation gives the combination method of edge computing and cloud storage, which could quickly deploy AI models in industrial scenarios, acquire the industrial process data via edge sensors, handle the complicated industrial issues with AI model and analytic modeling. The presentation gives the model transformation, pipeline organization, edge invocation, and demonstrates the application by UAC, fabricated decoration, and multi-channel HCI.

Assoc. Prof. Guangyu Gao
Beijing Institute of Technology, China

Guangyu Gao is an Associate Professor at the School of Computer Science and Technology, and served as the Deputy Director of the Institute of Data Science and Knowledge Engineering, at the Beijing Institute of Technology. He is also the secretary of the Secretariat of the China Engineering Education Accreditation Association (CEEAA), and the former Assistant Director of the Deputy Department of the Economic and Information Commission of Shunyi District, Beijing. He received his Ph.D. from the Beijing University of Posts and Telecommunications (BUPT) in 2013, and spent about one year at the National University of Singapore as a Joint-Ph.D. student from 2012 to 2013. He was awarded the IBM Visiting Scholarship in 2016, and the IBM Faculty Awards in 2016, 2017, and 2019.

His current research interests include Computer Vision, Deep Learning, and Multimedia Analysis. In recent years, he published more than 30 high-level academic papers (such as TPAMI, TCSVT, TMM, NeurIPS, CVPR, ACMMM, etc.) as well as a series of national invention patents. As the Principal Investigator, he is also undertaking 3 National Natural Science Funds, 4 University-Industry Collaborative Education Programs, and a number of School-Enterprise cooperation projects (with IBM, Taobao, etc.). He has successively won the Honorable Mention Award (second place) of the VisDrone Competition in IEEE ICCV-2019 (CCF-A), and the Nominate Video Award of the IEEE UIC-2015 Video Competition.

Speech Title: Addressing Few-shot Semantic Segmentation with Consideration of Feature Diversity
Fully-supervised semantic segmentation requires pixel-level annotations, which are costly and time-consuming, and also cannot well generalize to unknown categories. Thus, the Few-Shot Semantic Segmentation has been proposed to simulate human learning by inferring the semantic segmentation of a query image from a few annotated sample images of known classes. However, the most advanced few-shot semantic segmentation methods still have limited performance on public datasets due to typical problems such as ignoring the diversity of category features and instance features. To overcome these problems, we focus on the perspectives of category feature diversity and instance feature diversity, and propose the Hierarchical Context-agnostic Network and a Double Recalibration Network for few-shot semantic segmentation. On the one hand, the hierarchical context-agnostic network generates multi-scale feature representations for objects of different granularity in the annotated support images to provide diverse category guidance features. The background exclusion support module utilizes the context information in the background pixels to improve the model's discrimination ability for target objects and background regions. On the other hand, the double recalibration network considers the diversity of instance features by proposing a self-adaptive self-correction module and a cross-recalibration module for efficient fusion of prior information in the annotated support images to enrich the semantic features of query images. The two modules complement each other to form the double recalibration network. Finally, the proposed methods were evaluated on two public datasets, Pascal-5i and COCO-20i, and the results show significant performance improvements in few-shot semantic segmentation under diverse category and instance features, outperforming existing state-of-the-art methods.

Assoc. Prof. Dong Liang
Nanjing University of Aeronautics and Astronautics, China

Prof. Dong Liang received the B.S. degree in Telecommunication Engineering and the M.S. degree in Circuits and Systems from Lanzhou University, China, in 2008 and 2011, respectively. In 2015, he received Ph.D. at the research direction of System Information Science, the Graduate School of Information Science and Technology , Hokkaido University, Japan. He is currently an associate professor with the College of Computer Science and Technology, Nanjing University of Aeronautics and Astronautics. His research interests include model communication in pattern recognition and intelligent systems, optomechatronic and imaging systems. He was awarded the Best Student Paper Award from International Symposium on Optomechatronic Technology (ISOT) in 2013. He was awarded the Excellence Research Award from Hokkaido University in 2013. He has published several research papers including in Pattern Recognition, IEEE TIP/TNNLS/TGRS/TCSVT, and AAAI. He is an Organizing Committee member of the 2021 China Symposium on Machine Learning and Applications (MLA’21). He is a Program Committee member of the China Conference of Biometrics 2015-2022 (CCBR’15-22). He serves as a guest editor of MDPI Sensors. He has served as a reviewer for numerous international academic journals and conferences such as IEEE TIP/TNNLS/TGRS/TCSVT/TITS/JSTARS, and AAAI/IJCAI/ICML/ICPR/ICIP/ICASSP.

Speech Title: ALL-E: Aesthetics-guided Low-light Image Enhancement
Evaluating the performance of low-light image enhancement (LLE) is highly subjective, thus making integrating human preferences into image enhancement a necessity. Existing methods fail to consider this and present a series of potentially valid heuristic criteria for training enhancement models. In this talk, we introduce a new paradigm,i.e., aesthetics-guided low-light image enhancement (ALL-E), which introduces aesthetic preferences to LLE and motivates training in a reinforcement learning framework with an aesthetic reward. Each pixel, functioning as an agent, refines itself by recursive actions, i.e., its corresponding adjustment curve is estimated sequentially. Extensive experiments show that integrating aesthetic assessment improves both subjective experience and objective evaluation. Our results on various benchmarks demonstrate the superiority of ALL-E over state-of-the-art methods.

Assoc. Prof. Bo Li
Nanchang Hangkong University, China

Bo Li received the Ph.D. degree in 2008 in computational mathematics, Dalian University of Technology (DUT), Dalian, China. Now he is the professor in the School of Mathematics and Information Science of Nanchang Hangkong University. His current research interests include the areas of image processing and computer graphics. He has published more than 40 research papers in international journals (including IEEE TIP, IEEE TVCG, IEEE TMM) and conferences.

Speech Title: BBDM: Image-to-Image Translation with Brownian Bridge Diffusion Models
Image-to-image translation is an important and challenging problem in computer vision and image processing. In this talk, I will introduce a novel image-to-image translation method based on the Brownian Bridge Diffusion Model~(BBDM), which models image-to-image translation as a stochastic Brownian Bridge process, and learns the translation between two domains directly through the bidirectional diffusion process rather than a conditional generation process. To the best of our knowledge, it is the first work that proposes Brownian Bridge diffusion process for image-to-image translation. Experimental results on various benchmarks demonstrate that the proposed BBDM model achieves competitive performance through both visual inspection and measurable metrics.

Assoc. Prof. Hong Wu
University of Electronic Science and Technology of China, China

Dr. Hong Wu is an Associate Professor of Computer Science at University of Electronic Science and Technology of China (UESTC). He received his Ph.D. in pattern recognition and intelligent system from the National Laboratory of Pattern Recognition (NLPR) at the Institute of Automation (IA), Chinese Academy of Sciences (CAS), in 2004, and got his B.S. in Computer Science from the University of Science and Technology of China (USTC) in 1993. In May 2004, he joined NEC Labs China as an associate researcher undertaking research on web mining. In May 2006, he joined the University of Electronic Science and Technology of China (UESTC) as an associate professor. His current research interests are computer vision, pattern recognition and data mining.

Speech Title: Feature Aggregation and Refinement Network for 2D Anatomical Landmark Detection
Localization of anatomical landmarks is essential for clinical diagnosis, treatment planning, and research. In this paper, we propose a novel deep network, named feature aggregation and refinement network (FARNet), for the automatic detection of anatomical landmarks. To alleviate the problem of limited training data in the medical domain, our network adopts a deep network pre-trained on natural images as the backbone network and several popular networks have been compared. Our FARNet also includes a multiscale feature aggregation module for multi-scale feature fusion and a feature refinement module for highresolution heatmap regression. Coarse-to-fine supervisions are applied to the two modules to facilitate the end-to-end training. We further propose a novel loss function named Exponential Weighted Center loss for accurate heatmap regression, which focuses on the losses from the pixels near landmarks and suppresses the ones from far away. Our network has been evaluated on three publicly available anatomical landmark detection datasets, including cephalometric radiographs, hand radiographs, and spine radiographs, and achieves state-of-art performances on all three datasets.

Assoc. Prof. Xiaoli Zhang
Jilin University, China

Xiaoli Zhang is an associate professor in the Jilin University, China, where he is leading the research in the areas of multimodal image processing and deep learning in computer vision applications. He has published more than 60 research papers in international journals (including IEEE TIP, Bioinformatics, information sciences) and conferences. He is a senior member of CCF, member of CSIG and IEEE.

Speech Title: When Image Fusion Meets Deep Learning
Image fusion is an essential technique in computer vision that aims to combine information from multiple images to generate a single enhanced image with richer and more comprehensive details. Over the years, traditional image fusion methods have made significant contributions to various applications, such as medical imaging, remote sensing, and surveillance. However, with the advent of deep learning, the field of image fusion has witnessed a paradigm shift. This lecture explores the intersection of image fusion and deep learning, examining how deep learning techniques have revolutionized the field by enabling more accurate and robust fusion results. We delve into the fundamental concepts and challenges of both image fusion and deep learning, highlighting their individual strengths and limitations. Furthermore, we explore various deep learning architectures specifically designed for image fusion. We discuss how these architectures can effectively extract and fuse information from multiple input images, resulting in superior fusion outcomes.

Assoc. Prof. Jin Wang
Beijing University of Technology, China

Jin Wang is an Associate Professor with the Beijing Key Laboratory of Multimedia and Intelligent Software Technology, Faculty of Information Technology, Beijing University of Technology. He is also a researcher from Beijing Institute of Artificial Intelligence. He is an IEEE/ACM/CCF Member. His research areas include digital image processing, image/video coding, computer vision, and artificial intelligence. He authored more than 30 publications in journals and conferences, such as IEEE TIP, IEEE TMM, IEEE TCSVT, ACM MM, IEEE INFOCOM, IEEE DCC, and IEEE ICME. He holds more than 6 national invention patents. He has hosted more than 7 scientific research projects supported by National Natural Science Foundation and Beijing Municipal Natural Science Foundation, etc. He serves as a session chair of ACM MM 2021, a Technical Program Committee (TPC) member for ACM MM 2022, and a Program Committee (PC) member for AAAI 2022.

Speech Title: Multi-Scale Feedback Reconstruction for Guided Depth Map Super-Resolution
Guided depth map super-resolution (GDSR) is one of the mainstream methods in depth map super-resolution, as high-resolution color images can guide the reconstruction of the depth maps and are often easy to obtain. However, how to make full use of extracted guidance information of the color image to improve the depth map reconstruction remains a challenging problem. In this work, we first design a multi-scale feedback module (MF) that extracts multi-scale features and alleviates the information loss in network propagation. We further propose a novel multi-scale feedback network (MSF-Net) for guided depth map super-resolution, which can better extract and refine the features by sequentially joining MF blocks. Specifically, our MF block uses parallel sampling layers and feedback links between multiple time steps to better learn information at different scales. Moreover, an inter-scale attention module (IA) is proposed to adaptively select and fuse important features at different scales. Meanwhile, depth features and corresponding color features are interacted using cross-domain attention conciliation module (CAC) after each MF block. We evaluate the performance of our proposed method on both synthetic and real captured datasets. Extensive experimental results validate that the proposed method achieves state-of-the-art performance in both objective and subjective quality.

Assoc. Prof. András Horváth
Peter Pazmany Catholic Universtiy, Hungary

András Horváth is an Associate Professor at the Faculty of Information Technology and Bionics at Peter Pazmany Catholic University. His research is mostly focused on computer vision and artificial intelligence, especially on the efficient implementation of modern machine learning algorithms with emerging devices. He took part in the DARPA-UPSIDE (Unconventional Processing of Signals for Intelligent Data Exploitation) project between 2012 and 2018 in a consortium with Intel, MIT, which aimed the development of an object recognition pipeline with oscillatory based computing, implemented on emerging devices (e.g.: spin-torque and resonant body oscillators) and was involved in multiple international research grants sponsored by the European Union and ONR.

He is author or co-author of more than 50 publications which appeared in various international journals. He is an active Reviewer for various peer reviewed journal papers. (e.g.: IEEE Transaction on Signal Processing , IEEE Transactions on Circuits and System, etc.). He is a member of the IEEE Circuits and System and the IEEE Computational Intelligence Societies and the secretary of the Cellular Nanoscale Networks and Array Computing Technical Committee.

Speech Title: Preventing Adversarial Attacks in Deep Neural Networks
With the application of deep neural networks becoming mainstream in our everyday lives, questions about robustness and reliability of these networks are also becoming ever more important. Adversarial attacks reveal an important weakness of neural networks. Small perturbations of the input image easily cause misclassifications without changing the perception of a human observer. Among various attack methods, sticker-based attacks are the most threatening because they are robust enough for use in real world applications. These modifications cause an abrupt change on a small area of the input image and can invoke arbitrarily large activations in the network. In this talk I will introduce methods to detect and prevent adversarial attacks and yield safer applications in practice.

Assoc. Prof. Ruiheng Zhang
Beijing Institute of Technology, China

Ruiheng Zhang received the B.E. degree in 2014 from Beijing Institute of Technology, China, and the Dual-Ph.D. degree from University of Technology Sydney, Australia and Beijing Institute of Technology, China. He is currently an Associate Professor in Beijing Institute of Technology. He is the author of more than 30 research papers and one book, including Remote Sensing of Environment, IEEE TMM, ISPRS, Pattern Recognition, ICLR, IJCAI and so on. He is involved as a member of the Editorial Board of Frontiers in Robotics and AI, Artificial Intelligence and Applications. He has served as Co-Chair of IMASBD 2022, TPC of ICDIP 2022, VSIP 2022 and ICIVIS 2022. He is also a Reviewer for the ICLR, ACM MM, IJCAI, DICTA, IEEE TMM, IEEE TGRS, IEEE JSTARS, Pattern Recongition, Neurocomputing, Remote Sensing, Applied Science, Sensors, Electronics. His current research interests include deep learning, object understanding, and multi-modal remote sensing.

Speech Title: Infrared Target Detection with Limited Data
Target detection technology in infrared images under complex background has been widely used in many military and civil fields. In recent years, with its unique advantages in improving the efficiency of feature extraction and mining the hidden representation of data, deep learning has surpassed the traditional pattern recognition methods, providing a reliable way to achieve automatic target detection. However, the data requirement of deep learning restricts its performance on infrared target detection, where collecting a large number of infrared image datasets is time-cosuming and cost. How do we solve the contradiction between the demand of deep learning for large amounts of data and the lack of infrared image data? In addition, infrared images have the characteristics of low resolution, low contrast, low signal-to-noise ratio, fuzzy visual effect and lack of rich color information, which bring great challenges to the target detection task.

Assoc. Prof. Jingjing Zhang
China University of Geosciences (Wuhan), China

Jingjing Zhang received his B.S. degree in 2007 from China University of Geosciences, Wuhan, China. He obtained his Ph.D. degree in 2014 from the University of Chinese Academy of Sciences, China. From 2014 to now, he worked in the School of Automation, China University of Geosciences, Wuhan, China. He worked as a postdoctoral in the School of Energy and Power Engineering, Huazhong University of Science and Technology, from 2015 to 2017. He also worked as a visiting scholar at KU Leuven, Belgium, from 2017 to 2018. His research interests include augmented reality, computational photography, and infrared small target tracking.

Speech Title: Adaptive ROI Selection for Imaging Photoplethysmography
Region of interest (ROI) selection is a key process for remote heart rate measurement approach. Previous studies usually use a time window to make a goodness metric for ROI selection, which is computationally intensive and hard to apply for head rotation conditions. We proposed adaptive ROI selection methods to solve the problems. Results show that our proposed methods outperforms other methods.

Assoc. Prof. Yifan Zuo
Jiangxi University of Finance and Economics, China

Yifan Zuo, received the Ph.D. degree from the University of Technology Sydney, Ultimo, NSW, Australia, in 2018. He is currently an Associate Professor with the School of Information Management, Jiangxi University of Finance and Economics. His research interests include Image/Point Cloud Processing. The corresponding papers have been published in major international journals such as IEEE Transactions on Image Processing, IEEE Transactions on Circuits and Systems for Video Technology, IEEE Transactions on Multimedia.

Speech Title: Point Cloud Processing with Multimodality Guidance
Recently, there has been growing interest in point cloud processing with multimodality guidance. In this report, the speaker will introduce the principles and the advantages of multimodality guidance in point cloud processing and present the speaker's latest research achievements: (1) a multi-modal feature fusion method that enhances the distinctiveness of features for point cloud registration by leveraging texture information from images as guidance; (2) a method that combines pre-trained CLIP models with downstream tasks in point clouds. It effectively correlates features of point clouds and images in high-dimensional space, and improves the performance for point cloud downstream tasks.

Assoc. Prof. Ying Li
Nanjing Normal University, China

Ying Li is currently an Associate Professor in the School of Computer and Electronic Information / School of Artificial Intelligence, Nanjing Normal University (NNU). She received her Ph.D. degree from Dalian University of Technology (DUT) in 2019, advised by Prof. Xiangwei Kong. She was a visiting scholar at University of Texas at San Antonio, supervised by Prof. Qi Tian from Dec 2015 to Dec 2017. She worked as an intern in the Computer Vision group at the Noah’s Ark Lab of Huawei from Aug 2018 to May 2019. Her research interests are computer vision and pattern recognition. She is an Associate Editor (AE) of Pattern Recognition.

Speech Title: Challenges and New Trends for Fine-grained Image Retrieval
Fine-grained image retrieval is the task of retrieving images that belong to a specific subcategory within a larger category. This is a challenging task, as it requires the retrieval system to distinguish between subtle differences in visual features. This talk will discuss the challenges and new trends for fine-grained image retrieval, and introduce our recent works in related fields, including semantic retrieval, large-scale retrieval, multi-label pattern retrieval, multi-modal image-text retrieval, cross-domain person ReID and fine-grained action recognition.

Assoc. Prof. Dakshina Ranjan Kisku
National Institute of Technology Durgapur, India

Dakshina Ranjan Kisku is currently an Associate Professor in the Department of Computer Science and Engineering at National Institute of Technology Durgapur, India. He received his BCSE, MCSE and Ph.D. (Engineering) degrees in Computer Science and Engineering from Jadavpur University, India. Dr. Kisku is an active researcher in the area of biometrics, image processing, computer vision and machine learning. He was a Researcher in the Computer Vision Lab at the University of Sassari, Italy from March 2006 to March 2007. Prior to that, he was a Project Associate in Biometrics and Computer Vision Lab at Indian Institute of Technology Kanpur, India. He was a Postdoctoral Scientist in the Signal Processing Laboratory at Bahcesehir University, Turkey in 2012. Prior to joining NIT Durgapur, he was working as a faculty member for several years at Asansol Engineering College, India and Dr. B. C. Roy Engineering College, India. Dr. Kisku has more than 120 scientific publications to his credit published in refereed conferences, journals and edited books. He has co-authored and edited three books on biometrics and homeland security, and one book on medical biometrics. He has filed two patents in the area of Healthcare Informatics for publication. He is a recipient of IEI Young Engineers Award, IEI Excellence Award, Outstanding Scientist Award, Outstanding Reviewers Awards, IEEE Travel Award, IAPR Endorsement Scholarship, MIUR Research Fellowship, TUBITAK PDF Fellowship, Visvesvaraya Fellowship and many other accolades. Dr. Kisku is a Fellow of RSPH, UK and Senior Member of IEEE (USA). He regularly serves as a member of technical committee for many conferences and also reviewer for many refereed journals, conferences and books.

Assoc. Prof. Chunzhi Li
Huzhou University, China

Chunzhi Li received the M.Sc. and Ph.D. degrees respectively in the Department of Computer Science from Wuyi University, Jiangmen, China and in the Department of Computer Science and Technology, East China Normal University, Shanghai, China, in 2007 and 2014. She is currently an associate professor with the Department of School of Information Engineering, Huzhou University, Huzhou City, Zhejiang Province, China, also a visiting scholar with the College of computer science and technology, Zhejiang University, Zhejiang Province, China. She has been a visiting associate professor in 2019-2020 with the Department of Electrical and Electronic Engineering, the University of Hong Kong, Hong Kong. Her main research area is Remote Sensing Image Processing and has published more than 20 papers on “remote sensing” in the excellent journals, such as IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING.

Speech Title: Exploring Oblique Rotation Factor to Restructure Deep Hyperspectral Image Classification
Factor analysis (FA) is commonly used in fields such as economics and now being introduced as a new tool on dimensionality reduction (DR) for hyperspectral image classification (HSIC), but FA usually employed orthogonal rotation to directly maximize the separation among factors, which would oversimplify the relationships between variables and factors, worse still, the orthogonal rotation often distorts the true relationships between underlying traits in real life and can not always accurately represent these relationships. To this end, this letter proposes a DR algorithm about FA based on oblique rotation Oblimax to improve HSIC. Firstly, the common factors are extracted from the hyperspectral data to form a factor loading matrix which will be obliquely rotated, then its factor score is estimated to obtain the eigen dimensions for the hyperspectral data, thus realizing DR. On the basis of the successful DR, a deep classifier is constructed, specially, a double-branch structure about 3 dimensional-convolutional neural networks (3D-CNN) with different sizes is restructured to extract multi-scale spatialspectral features, and early fusion is performed on the features, then 2 dimensional-convolutional neural networks (2D-CNN) is restructured to reduce the computational complexity and learns more spatial features. Finally, the accuracy of the proposed algorithm on the datasets Indian Pines, Kennedy Space Center and Muufl Gulfport, respectively achieves 99.78%, 99.95% and 95.57%. It shows that the proposed algorithm in this letter has advantages in improving the classification accuracy and reducing the complexity of computation.

Assoc. Prof. Yukinobu Miyamoto
Kobe Gakuin University, Japan

MIYAMOTO Yukinobu received the B.E., M.E. and Ph.D. degrees in computer science from Kobe University. He had studied artificial intelligence, especially machine learning, pattern recognition and evolutionary computation. He is currently an Associate Professor at Faculty of Business Administration, Kobe Gakuin University. He is concurrently a Professor at Division of Information Systems, Graduate School of Information Technology, Kobe Institute of Computing. He is also a Professional Engineer in Information Engineering (P.E.), and a Certified Information Technology Professional (CITP).

Speech Title: An Approach for Rotation and Scale Robust Texture Recognition Using Wavelet Transform and Moment Features
Texture is one of the key characteristics used to identify objects or regions of interest in an image. One of the difficulties in texture analysis has been the lack of appropriate tools for characterization. Previous texture analysis methods have been based on modified wavelet transforms, called tree-structured wavelet transforms, and what are classified as neural networks. However, it is difficult to classify textures of different scales and rotations using only these methods. We propose texture recognition based on moment features. We compute the gray-level density function of each image and compare its histogram, which represents its shading. Histogram comparison is a commonly used method for indexing images and expressing the histogram in terms of its moments reduces the complexity of the method. For the computation of moment features, the wavelet transform corresponding to each node is used. For classification, each moment feature is fed to a neural network. This method allows the recognition of textures of different scales and rotations and improves the recognition rate.

Assoc. Prof. Indrarini Dyah Irawati
Telkom University, Indonesia

Indrarini Dyah Irawati, obtained B.E and M.E degree in Electrical Engineering at Telkom University, Bandung, Indonesia and doctoral degree in the School of Electrical and Information Engineering, Institute of Technology Bandung. She joined School of Applied Science, Telkom University as an Instructor (2007-2013), Assistant Professor (2014-2018), Associate Professor (2019-present). Her main research interests are in the areas of compressive sensing, watermarking, signal processing, computer network, and IoT. She is currently a member of the Association for Computing Machinery (ACM), the International Association of Engineers (IAENG), and IEEE. She has published more than 25 research papers-mostly the top ranked journal Scopus Indexed by IEEE, Springer, Elsevier, and 20 Conference papers-mostly the top ranked international conference, 3 books, 1 book chapter, 15 Intellectual property rights, and 1 patent.

Speech Title: Review on Enhanced Diagonal Minimization (EDMIN) for Sparse Two-Dimensional Data
Traffic matrix (TM) is important information that is used by researchers to develop research in the context of network monitoring, network capacity planning, and network maintenance. TM is obtained from measurements directly on the router in the network. Compressive Sensing (CS) provides an alternative solution to overcome problems on the internet network as stated earlier, reducing the number of traffic measurement samples on the network by compression. A novel Enhanced Diagonal Minimization (EDMIN) algorithm used for reconstructing two-dimensional data based on compressive sensing (CS). EDMIN finds solutions with a global approach to obtain SV diagonal matrix so that it is superior to other conventional algorithms.