4th August 2021 (Wednesday)
9.00 A.M - 16.30 P.M
IEEE Signal Processing Society Malaysia Chapter and Universiti Kuala Lumpur Malaysian Institute of Information Technology are jointly organizing the 12th Symposium on Image Processing, Image Analysis and Real-Time Imaging (IPIARTI 2021).
This FREE annual event, open to all members and non-members of the IEEE, is organized with the objectives of:
09.00 - 09.30 : Registration
09.30 - 09.45 : Welcome address
09.45 - 10.30 : Keynote Speech #1 - Deep Learning Applications in Image-based COVID-19
Assoc. Prof. Dr. Mohd Asyraf Zulkifley, UKM
10.30 - 10.50 : Break
10.50 - 11.40 : Keynote Speech #2 - Preventing Cable-Cuts Caused by Construction
Activities: An IoT and
AI Image Processing Application
Dr Qazi Mamoon Ashraf, TM Research & Development
11.40 - 12.30 : Keynote Speech #3 - Dense Crowd Count Estimation Research Based On CSRNET
Dr Megat Norulazmi Megat Mohd Noor, UniKL MIIT
12.30 - 13.00 : Membership drive and Announcements
13.00 - 14.00 : Lunch Break
14.00 - 16.40 : Technical Presentations
16.40 - 17.00 : Certificate Presentation & Closing
Prospective presenters are invited to submit a one-page abstract of their work and will be given a certificate of appreciation as invited speaker.
Bio: Mohd Asyraf Zulkifley received his Bachelor Degree of Engineering (Mechatronics) from International Islamic University Malaysia in 2008 and Ph.D. (Electrical and Electronics) from The University of Melbourne in 2012. In 2014 and 2015, he was a visiting academic and visiting researcher at The University of Melbourne and MIMOS, respectively. He was then appointed as an associate professor at the Universiti Kebangsaan Malaysia in 2016, where he was later attached to the Department of Computer Science, the University of Oxford for two years as a sponsored researcher, working with Prof. Niki Trigoni. Recently, he received the Turkish Burslari Fellowship award for short attachment at the Sabanci University, Turkey. His main research interest is deep learning applications in computer vision.
Title: Deep Learning Applications in Image-based COVID-19 Screening and Diagnosis.
Since the first COVID-19 case was spotted in Wuhan, China at the end of 2019, the pandemic has taken more than 4 million lives all over the world. COVID-19 is a contagious disease that can be easily spread to others even with minimal physical contact. Hence, one of the best strategies to curb the spread of this disease is through mass screening tests, so that the patients can be quarantined at the dedicated centers as early as possible. However, the widely used Reverse Transcription Polymerase Chain Reaction (RT-PCR) swab test to diagnose the COVID-19 disease, generally requires more than 24 hours lead time before the conclusive report can be obtained. This considerably long period of waiting time may cause the disease to be spread to the others already. Therefore, a faster medical image-based approach has been proposed to screen the disease immediately, such as through x-ray imaging and CT scan. Yet, it is a tedious and laborious effort to identify all those tiny features to represent and confirm the possibility of COVID-19 disease. To overcome this challenge, several deep learning techniques have been used to automate the detection of this disease. In this talk, I will explore several state-of-the-art techniques that were recently developed to screen the COVID-19 disease.
TM Research & Development
Bio: Qazi Mamoon Ashraf is a product researcher with Telekom Malaysia Research and Development, where he is involved in product development in emerging technologies especially in Internet of Things (IoT) and Artificial Intelligence (AI) applications. He received his Ph.D. degree in Computer Engineering from University Islam Antarabangsa, Malaysia. Previously, he was a Research Assistant with the Wireless Communication Division in MIMOS, Malaysia. He has authored ten filed patents in wireless sensor telecommunication, and seventeen international research publications. He is also a member of IoT standardization group at Malaysian Technical Standards Forum Bhd (MTSFB) under Malaysian Communications and Multimedia Commission (MCMC). His research interests include IoT, Internet protocols, energy efficiency, autonomic computing, M2M communication systems, and ubiquitous networks with a focus on telecommunication industry adoption.
Title: Preventing Cable-Cuts Caused by Construction Activities: An IoT and AI Image Processing Application.
The keynote will focus on image processing technology and supervised labelling in image recognition and its application to solve a real-life telecommunication industry problem. It will introduce an inhouse developed solution called Patrol. Suitable for utility companies and government agencies, Patrol prevents utility service disruptions to end users due to construction threats on Fiber Infrastructure via real-time alerts. Patrol combines user-friendly Artificial Intelligence (AI) & supervised machine learning algorithms, high-quality video data, and reliable methodology to quickly and accurately assess construction risk automatically. Subsequently, operational team can quickly create thorough and accurate alerts about construction conditions. Patrol has shown promising results and it is expected to provide a solid base for the operationalizing automatic acquisition of construction risk information.
Bio: Megat Norulazmi Megat Mohamed Noor is a Senior Lecturer at the Malaysian Institute of Information System, Universiti Kuala Lumpur. He graduated in Bachelor of Science in Engineering from University of Kagoshima, Japan and completed a Master of Information Technology at Open University Malaysia. He was awarded a PhD from Universiti Utara Malaysia in Information Technology. His recent research interests are Deep Learning, Cloud Computing, and IoT. Currently he is the Principal Investigator of MIIT’s Centre of SDN/NFV & IoT. He is an active fellowships member of Okinawa Open Laboratory, Japan
Title: Dense Crowd Count Estimation Research Based On CSRNET.
Crowd counting is an active area of research and has seen several developments since the advent of deep learning. Crowd counting is a technique to estimate the number of people in an image or a video. However, if there are too many people crammed in the image, it makes a huge task for a human brain to accurately predict the right number to estimate. Nowadays, the deep learning CNN-based model is the recent state of the art technic that able to produce the best crowd count estimation performance. In today's talk, I will share our ongoing research group activities on this topic which is supported by the Co-PI RDO grant of the Saudi Arabia Ministry of Higher Education. Initially, I will explain the fundamental background of the crowd count estimation based on CSRNET model. After that, I will deep dive into several proposed technics related to data preparation and convolutional neural network architecture that to be used in this research to achieve the objectives set by our stakeholders.