Dr Mustansar Fiaz, IBM Research,a United Arab Emirates
Passionate about computer vision and deep learning, Mustansar Fiaz specializes in remote sensing, person search, medical image segmentation, visual object tracking, and multi-modal analysis. With a Ph.D. from Kyungpook National University, he has over three years of post-PhD experience. Mustansar currently leads innovative research at IBM Research in Abu Dhabi, developing advanced models for remote sensing applications. His prior roles include research associate at MBZUAI and senior AI software engineer at Tricubics, Seoul. His notable projects and numerous awards reflect his contributions to the field. 🌍🖥️🔬🛰️
Publication profile
google scholar
Education
Dr. Mustansar Fiaz, completed a Ph.D. in Computer Science and Engineering from Kyungpook National University, Daegu, S. Korea (Mar. 2016 – Feb. 2021) with a GPA of 4.18/4.5. Their thesis focused on “Robust Object Tracking and Segmentation Using Siamese Networks.” They earned a Master’s in Engineering from Sejong University, Seoul, S. Korea (Mar. 2014 – Feb. 2016) with a GPA of 4.39/4.5, researching the “Space Knowledge Information Process (SKIP) Tool for Multi-Dimensional Data Analysis.” They hold a BS in Computer and Information Sciences from PIEAS, Islamabad, Pakistan (Mar. 2007 – Aug. 2011) with a GPA of 3.01/4.0, where they worked on “Medical Image Segmentation using Statistical and Transform Methods.” 🎓💻📊🖼️
Experience
Currently a Staff Research Scientist at IBM Research in Abu Dhabi, UAE (Oct. 2023 – Present), where I specialize in remote sensing applications and develop Visual-Language foundation models for remote sensing. Previously, He was a Research Associate at the Intelligent Visual Analytics Lab (IVAL) at MBZUAI (Aug. 2021 – Oct. 2023), focusing on computer vision tasks, writing research proposals, supervising CV701 labs, and co-supervising MS and Ph.D. students. I also served as a Senior AI Software Engineer at Tricubics, Seoul (Mar. 2021 – Jul. 2021), leading AI and computer vision R&D for AI-based unmanned stores, and as a part-time CTO at Ujura (CATMOS), Seoul (Mar. 2021 -May. 2021), developing AI health information systems for pets, especially cats. 🛰️📊🐱
Noteable Projects
Mustansar Fiaz, a researcher and developer at MBZUAI in Abu Dhabi since August 2021, specializes in remote sensing, medical image segmentation, and person search. His work in remote sensing involves developing AI-based algorithms to detect semantic changes in man-made facilities while ignoring noisy changes. In medical image segmentation, he creates CNN and transformer-based models for 2D and 3D segmentation tasks, including cell, multi-organ, polyp, and cardiac MRI scans. For person search, he develops methods to detect and localize individuals in images from multiple cameras. Previously, he worked at KNU in Daegu, South Korea, focusing on video object tracking and segmentation using deep learning techniques. 📊📸🏥🛰️
Awards
Recipient of the Best Presentation Award and Best Paper Award at IW-FCV 2022 🏆, with additional accolades including the Best Student Paper Award at IW-FCV 2020 and the Outstanding Research CSE Thesis Award in 2021 🎓. Honored with the ACM SIGAPP Student Travel Award for ACM SAC 2019 ✈️. Served as Joint Secretary and Secretary Information for PSAK between 2017-2019 📜. Awarded the KNU International Graduate Scholarship (KINGS) for PhD studies (2016-2021) 🎓, Sejong University’s fully funded MS scholarship (2014-2016) 📚, and PIEAS Fellowship for BS studies (2007-2011) 🏅.
Research focus
Based on the publications provided, the research focus of this person centers on visual object tracking and remote sensing. They are involved in developing advanced methods for tracking objects in noisy and cluttered environments, including techniques like channel-spatial attention learning and deep Siamese networks. Their work also explores Medical image segmentation and gender identification using signal processing techniques. Recent studies highlight the use of transformers and Siamese networks for robust tracking and change detection in remote sensing. The emphasis is on improving tracking accuracy and robustness through novel algorithms and model architectures. 🌟📊🔍📷
Publication top notes
Handcrafted and deep trackers: Recent visual object tracking approaches and trends
Tracking noisy targets: A review of recent object tracking approaches
Gender identification using mfcc for telephone applications-a comparative study
Efficient visual tracking with stacked channel-spatial attention learning
Medical image segmentation using h-minima transform and region merging technique
Remote sensing change detection with transformers trained from scratch
Learning soft mask based feature fusion with channel and spatial attention for robust visual object tracking
Deep siamese networks toward robust visual tracking
Sat: Scale-augmented transformer for person search
Improving object tracking by added noise and channel attention