Registration (10:00 - 10:30)
10:30 - 10:45
Host: Prof. Naoya Iwamoto, Kagawa Kosen
Opening Address by President Nobuo Araki, Kagawa Kosen
Opening Address by Prof. Ming-Der Shieh, National Cheng Kung University
Program Overview
10:45 - 11:35
Host: Prof. Naoya Iwamoto, Kagawa Kosen
Title: Next-Generation Regional Public Transport - Autonomous-Driving Bus Business in Japan -
Speaker: Mr. Hajime MIYAZAKI, NTT WEST
Abstract: Japan’s regional public transportation faces growing challenges due to population decline, aging demographics, and driver shortages. In response, Japan’s autonomous bus market is accelerating, supported by government initiatives that emphasize Level 4 (L4) autonomous operation. This presentation outlines key market and policy trends, NTT West’s initiatives and deployment experiences, and the basic mechanisms of autonomous driving systems. It also discusses how L4 autonomous buses can enable scalable, resilient, and sustainable regional transportation.
Bio: Hajime MIYAZAKI is currently Director of Business Development at NTT West. After joining NTT, He worked on ISP, Data Center and Submarine Cable business, with overseas assignments in the UK and Thailand, and later served as Director of Asian strategy at NTT Communications (now NTT DATA). He has extensive experience in autonomous mobility and AI, including serving as Director of Business Development at Morpho, an AI startup from The University of Tokyo. He also serves as a board member of Navya Mobility, a France-based autonomous driving R&D company, and holds an M.S. degree from the Graduate School of Engineering Science, The University of Osaka.
Break (11:35 - 11:45)
11:45 - 12:35
Session Chair: Prof. Kohei Ishii, Kagawa Kosen
11:45
Title: AI-Driven Digital Twins for Smart Manufacturing and Robotic Systems
Speaker: Prof. Ming-Der Shieh, National Cheng Kung University
Abstract: The concept of digital twins has been proposed for more than a decade. With the rapid advancement of AI technologies and capabilities, applications of digital twins have flourished significantly. Various development environments and platforms—such as NVIDIA’s Omniverse, Cosmos, and Isaac Sim—have further created opportunities for rapid deployment. Applications such as embodied AI and physical AI have emerged as prominent research areas and present significant market opportunities nowadays. In this talk, I will use the digital twin systems developed at ITRI over the past few years as examples to explore related technology development and applications.
Bio: Ming-Der Shieh received his B.S. degree in electrical engineering from National Cheng Kung University, Taiwan, in 1984, his M.S. degree in electronic engineering from National Chiao Tung University, Taiwan, in 1986, and his Ph.D. degree in electrical engineering from Michigan State University, USA, in 1993. Since 2002, he has been with the Department of Electrical Engineering, National Cheng Kung University, where he is currently a professor. He has also served as Chief Technology Officer of the Electronic and Optoelectronic System Research Laboratories at Industrial Technology Research Institute (ITRI) since 2022. His research interests include VLSI design and testing, VLSI for signal processing, and digital communication.
12:10
Title: Applications of brain–computer interfaces: Brain-controlled robotic arm and spatial target selection
Speaker: Prof. Akinari Onishi, Kagawa Kosen
Abstract: Brain-computer interfaces (BCIs) enable users to control various robots such as robotic arms via brain signals. Since BCIs are largely independent of muscle activity, they can support people with disabilities to control external devices. This talk presents our group’s recent robotic applications of BCIs including a camera robot, an autonomous wheelchair, and an autonomous robotic arm controlled by the spatial target selection using BCIs.
Bio: Akinari Onishi received his Ph.D. degree from Kyushu Institute of Technology in 2015. He is currently a Senior Lecturer at the National Institute of Technology, Kagawa College (Kagawa KOSEN). His research interests include brain-computer interfaces, biomedical engineering, assistive technologies, and engineering education.
Group Photo & Lunch Break (12:35 - 13:50)
13:50 - 15:05
Session Chair: Prof. Chih-Tsun Huang, National Tsing Hua University
13:50
Title: Vision-Language Models for Rewards in Reinforcement Learning
Speaker: Prof. Min-Chun Anita Hu, National Tsing Hua University
Abstract: While reinforcement learning has seen great success, engineering reward functions for abstract tasks remains a persistent challenge. This talk introduces a novel framework that leverages the zero-shot capabilities of Vision-Language Models to provide reward signals. By grounding an agent's visual experiences in natural language, we demonstrate how reward functions can be effectively learned from VLM-generated preferences in complex, open-ended settings.
Bio: Dr. Min-Chun Hu earned her Ph.D. from the Graduate Institute of Networking and Multimedia at National Taiwan University in 2011. Following a postdoctoral fellowship at the Research Center for Information Technology Innovation, Academia Sinica, she held professorships at National Cheng Kung University before joining the Department of Computer Science at National Tsing Hua University, where she currently serves as a Professor.
Dr. Hu has received prestigious accolades including the NSTC Ta-You Wu Memorial Award (2023), the CES Innovation Award (2023), and the Google Research exploreCSR Award (2021–2024). Her core research encompasses computer vision, robotic AI, multimedia retrieval, and immersive technologies (AR/VR). As a lifelong basketball enthusiast, she bridges the gap between technical engineering and the nuanced demands of the sports and arts industries. This is evidenced by her leadership as the former Deputy Executive Director of the Taiwan Institute of Sports Science and her role as co-founder of NeuinX, a startup dedicated to AI-driven sports analytics.
14:15
Title: Deep Learning Assisted 5G/6G Wireless Communication Systems
Speaker: Prof. Taisei Urakami, Kagawa Kosen
Abstract: In fifth- and sixth-generation (5G/6G) wireless communication systems, millimeter-wave (mmWave) band is widely adopted to achieve broader bandwidth, higher data rates, and lower latency. However, due to the highly directional propagation and significant path loss of mmWave signals, coverage becomes severely limited, especially in non-line-of-sight (NLOS) regions. Massive multiple-input multiple-output (mMIMO) and reconfigurable intelligent surface (RIS) technologies have been introduced to mitigate these dead zones by leveraging beamforming techniques based on channel state information (CSI). In particular, the rapid advancement of deep learning (DL) has further enhanced beamforming performance in both mMIMO and RIS systems. In this presentation, we introduce our recent works on DL-based beamforming schemes.
Bio: Taisei Urakami received the B.S. degree from National Institute of Technology, Kagawa College, Kagawa, Japan in 2022, and the M.S. and Ph.D. degrees from Nara Institute of Science and Technology, Nara, Japan, in 2023 and 2025, respectively. He is currently an Assistant Professor with National Institute of Technology, Kagawa College, Kagawa, Japan. He is a member of IEEE and IEICE. His research interests include deep learning-based beamforming for reconfigurable intelligent surface-aided mmWave systems, metasurface reflector, and microwave circuits.
14:40
Title: Fingerspelling Recognition Using Fingernail Strain
Speaker: Prof. Kohei Ishii, Kagawa Kosen
Abstract: Strain is generated in the nail due to the movement and deformation of the surrounding tissues during finger motion. By measuring this strain and using it as training data, it may be possible to estimate and recognize finger movements. In the experiment, machine learning methods were used to estimate finger spelling from nail strain and fingertip posture information, and their accuracy was evaluated. The results confirmed that adding nail strain information improves estimation accuracy compared to using only fingertip posture information (quaternions).
Bio: He received a Ph.D. degree in medicine from the University of Tokyo in 2013, and is presently an associate professor at Department of Electro-Mechanical Systems Engineering, National Institute of Technology, Kagawa College. His research interests are biomedical engineering, in particular wearable sensors and devices working on nail surface. He is a member of Japanese Society for Medical and Biological Engineering (JSMBE) and IEEE Engineering in Medicine
and Biology Society (EMBS).
Break (15:05 - 15:25)
15:25 - 16:40
Session Chair: Prof. Yukikazu Murakami, Kagawa Kosen
15:25
Title: Multimodal AI-Based Predictive Modeling for Campus Environmental Risk Assessment Using Limited Heterogeneous Data
Speaker: Prof. Yi-Yu Alan Hsu, National Cheng Kung University
Abstract: Ensuring campus safety is a multi-dimensional challenge that requires monitoring physical infrastructure alongside social and behavioral indicators. However, the development of robust AI-driven predictive models for this task is frequently hampered by the "small data" problem, where high-quality, campus-specific imagery is scarce. This research presents a multimodal deep learning framework to predict environmental risks by fusing a limited set of campus photos with structured risk assessment scores. To mitigate the risk of overfitting in data-constrained scenarios, the framework utilizes vision language models with pre-trained ResNet50 architectures and employs advanced data augmentation techniques including affine transformations and neural filters to expand the training distribution. We implement an intermediate (feature-level) fusion strategy to effectively align visual patterns from photos with structured numerical data, enabling the model to learn complex inter-modal relationships. The proposed system facilitates a strategic shift from post-incident response to proactive early warning in educational settings, significantly enhancing the efficiency and precision of campus security governance.
Bio: Dr. Yi-Yu Alan Hsu is an assistant professor at Miin Wu School of Computing, NCKU, Taiwan. He received a Ph.D. degree in computer science and information engineering from NCKU in 2015. His research interests are artificial intelligence, data mining, pattern recognition, and natural language processing.
15:50
Title: Automated Recycling Waste Classification Using YOLOv8 and Content-Based Image Retrieval
Speaker: Prof. Jeng-Han Roger Li, Southern Taiwan University of Science and Technology
Abstract: This study aims to improve recycling practices by leveraging modern machine learning and web search capabilities to develop a system that effectively adjusts to changes in waste materials. Utilizing a combination of image web search and object detection algorithms, the developed model accurately identifies objects from 10 classes and can expand to include new classes by acquiring new images. The YOLOv8 deep learning model uses a dataset containing over 69,919 waste images across 10 categories and achieved an impressive accuracy of 99.5% over 500 epochs. The images unknown to the model will be captured and processed with the Content-Based Image Retrieval (CBIR) algorithm to download images from the internet and use them as samples for model training. This dynamic system continuously evolves as added items are classified and integrated, enhancing overall accuracy and mitigating misclassification of recyclable and non-recyclable materials.
Bio: Prof. Jeng-Han Li is an Assistant Professor in the Department of Electrical Engineering at Southern Taiwan University of Science and Technology. He received his Ph.D. degree in Electrical Engineering from National Cheng Kung University in 2003. His expertise spans both academia and industry. Between 2014 and 2016, Prof. Li contributed significantly as a project manager at the Productibility Promotion Office of MOEA in Taiwan. Since joining STUST in 2016, he has excelled in teaching and actively participated in the robotics community, serving on the competition rules review and referee committees for the prestigious Top International Robotic Tournament since 2018. Prof. Li's work has earned him numerous awards in the National College Creative Design and Production Competition, the DSP Creative Design Contest, and the TIRT Robot Competition.
16:15
Title: FST based automatic word segmentation and part-of-speech tagging for Machine translation (a case-study on Myanmar)
Speaker: Prof. Tin Htay Hlaing, Kagawa Kosen
Abstract: Machine translation of Asian languages especially less privileged language, Myanmar has many areas to improve. In the meantime, the time and cost to build a single and standard corpus for Myanmar language is also challenging task. This research aims to develop an automatic word segmentation without the use of corpus. In other words, the segmentation is done based on the types of prepositions in formal( i.e not conversational) Myanmar sentences. In this research, FST is utilized in segmentation and part of speech tagging. Though the experiments are underway to prove the accuracy, the proposed method is shared in this workshop.
Bio: Tin Htay Hlaing completed her first master degree in computer science at University of Computer studies, Yangon in 2003. In 2009, she continued her studies in Japan and graduated in 2011 and 2014 for Master of Engineering and Doctor of Engineering respectively.
16:40 - 16:50
Host: Prof. Naoya Iwamoto, Kagawa Kosen
Closing Remarks by Prof. Yukinori Misaki, Kagawa Kosen
Banquet