Dr. Po-Tsun (Paul) Kuo
National Yang Ming Chiao Tung University

Speech Title: Deploying Pattern Recognition and Computer Vision Systems in Manufacturing and Medical Imaging: Practical Lessons from Industry
Abstract: Recent advances in pattern recognition and computer vision have demonstrated impressive performance in laboratory and benchmark settings; however, deploying these models reliably in real industrial environments remains a significant challenge. This invited talk shares practical lessons learned from deploying AI systems in both electronic manufacturing and medical imaging applications, including automated optical inspection (AOI), workplace safety surveillance, and medical image analysis for CT, X-ray, and ultrasound. The talk focuses on best practices that enable smooth transition from research prototypes to long-term industrial operation. Key topics include data preparation strategy, training strategy and model lifecycle management, with particular emphasis on retraining mechanisms to sustain performance under changing production and clinical conditions. In addition, the role of an AI platform in supporting scalable deployment—covering data versioning, model monitoring, validation, and continuous improvement—will be discussed. Through real-world case studies, this talk highlights how systematic integration of these strategies can significantly reduce deployment risk, shorten time-to-value, and ensure stable AI performance in long-run industrial and medical applications.

Bio-Sketch: Dr. Po-Tsun (Paul) Kuo received his B.Eng. and M.Eng. degrees in Electrical and Electronic Engineering from the University of Auckland, New Zealand, in 2001 and 2003, respectively, and his Ph.D. degree in Electronics and Electrical Engineering from the University of Edinburgh, UK, in 2006. From 2006 to 2009, he was a postdoctoral researcher at the Digital Image Research Centre, Kingston University, London, where his work focused on image analysis and computer vision. He then worked as a computer vision researcher at Imperial Innovations, affiliated with Imperial College London. Between 2010 and 2015, he held technical roles in Automatic Optical Inspection (AOI) companies in Taiwan, developing industrial computer vision solutions for electronic manufacturing. In 2015, Dr. Kuo joined Advantech Co., Ltd. as a Senior Technical Manager, leading the AI Vision team within the Cloud Computing Division and driving the productization of AI technologies for industrial applications. Since April 2025, he has been with Delta Electronics, Inc. as a Senior Technical Manager in the Delta Smart Manufacturing Department, focusing on AI-driven optimization of design, manufacturing, production, and inspection processes. He has also held a joint appointment as an Associate Professor at National Yang Ming Chiao Tung University since 2020. His research interests include industrial and medical computer vision, deep learning, and long-term AI system deployment.

Dr. Priyanka Harjule
Malaviya National Institute of Technology (MNIT) Jaipur, India

Speech Title: Detecting Silicosis on X-rays Using a Deep Learning Model Based on a Fractional Optimizer
Abstract: Among the most prevalent harmful respiratory diseases in Asia is silicosis. Silicosis is an irreversible fibrotic lung disease arising from prolonged exposure to respirable crystalline silica dust. Radiologists expertise is essential in identifying silicosis on X-ray, resulting in delayed detection and treatment in mining areas. Machine learning-based computeraided detection approaches are being evaluated as a potential solution to this concern. However, the lack of extensive, publicly accessible silicosis image databases complicates the development of precise deep-learning models for silicosis detection via X-rays. Screening of silicosis exhibits complex characteristics reliant upon lung architecture, typically revealing regional similarities, particularly across Asian populations. Still, no advanced models have been introduced to eIectively capture these features to enable accurate diagnoses. In my talk, I will talk about a deep learning model with fractional optimizer using a CNN-based transfer learning approach to detect silicosis. The proposed method uses fractional momentum gradient descent optimizer incorporating Caputo derivative, which modify the learning potential of the network with one more parameter $\alpha$. The convergence of this optimizer towards minimum has been mathematically proven. In my talk, I will also show that our proposed model has enhanced learning capacity, potentially leading to remarkable performance improvements comparative to others. The experiments comprises two datasets one is pneumonia which is a popular dataset available publicly and the second dataset is composed of a combination of Indian and Chinese lung images, contains both normal and silicosis aIected lungs. The results shows that proposed model achieves better performance with fractional orders greater than $\alpha = 1$, and the optimal $\alpha$ order is detected as $\alpha =1.9$.

Bio-Sketch: Dr. Priyanka Harjule is currently a joint faculty member in the Departments of Mathematics and the Department of Artificial Intelligence & Data Engineering at Malaviya National Institute of Technology (MNIT) Jaipur, India. Her research interests lie at the intersection of Applied Mathematics, Machine Learning, Optimization, and Fractional Calculus. She has authored over 30 research publications in reputed journals and international conferences. Dr. Harjule's recent research includes the design of fractional optimizers for deep neural networks and development of fractional derivative-based filters for texture enhancement in medical imaging. Prior to joining MNIT, she served as Assistant Professor at the Indian Institute of Information Technology (IIIT) Kota, where she also founded and coordinated the e-Learning and Data Analytics Lab (e-LDA). Dr. Harjule has been actively involved in national and international research collaborations, and has participated in several government-sponsored research projects. Her contributions span both theoretical and applied aspects of mathematical modelling and computation, with a strong emphasis on interdisciplinary applications in engineering and healthcare.

 

coming more......