Dr. Terence Sim
School of Computing, National University of Singapore
Associate Professor and Vice Dean of Communications
Dress-up for CS1231 lecture
The year I graduated 3 PhDs!
Recognize faces even if they are under heavy makeup, by using motion signatures.
About Dr. Terence Sim
Explain. Demonstrate. Experiment. Inspire.
The above sums up my teaching and research philosophy. Over the years, I had the pleasure of teaching many courses - Introductory Programming, Computer Vision, Visual Effects, Multimedia - and interacting with many talented students. I'm currently teaching Discrete Structures, and Biometrics Authentication.
For research, I explore several areas related to Visual Computing: Facial image analysis, Multimodal biometrics, Facial rendering, Computational photography, Continuous authentication, Music transcription, to name a few. I combine machine learning with physics-based modeling and graphics rendering to tackle the challenges in research.
I also provide consultancy in biometrics.
Learning Controllable Face Generator from Disjoint Datasets
CAIP, September 2019
Jing Li, Yongkang Wong, Terence Sim
Disentangling semantic features (such as illumination, pose, and identity) from a set of face images very difficult, but immensely useful. The problem is compounded because large fully labeled datasets are not available. This paper proposes a bridging dataset to overcome this problem.
Understanding Humans in Crowded Scenes: Multi-Human Parsing
ACM MM 2018, Best Student Paper Award
Jian Zhao, Jianshu Li, Yu Cheng, Terence Sim, Shuicheng Yan, Jiashi Feng
Multi-human parsing is about segmenting body parts in a scene with multiple people, and associating each part with the correct person. This is useful for understanding the scene.
Task Relation Networks
Jianshu Li, Pan Zhou, Yunpeng Chen, Jian Zhao, Sujoy Roy, Shuicheng Yan, Jiashi Feng, Terence Sim
Multi-task learning is a popular strategy to enable a network to simultaneously learn several tasks in order to benefit from the shared knowledge in the related tasks. But how do you choose what tasks to learn from? This paper proposes a novel metric to measure task similarity, and exploits it in the network architecture.
Profiling Biometric Authentication on Mobile Devices
Sanka Rasnayaka, Sanjay Saha, Terence Sim
Smartphones routinely use multiple biometrics to authenticate users (eg. face, voice, gait). But what is the impact on battery power and device memory? This paper is the first to systematically profile the resource consumption of biometric verification systems.
Courses and Talks
CS2030S Programming Methodology II
Semester 1, AY2021-22
This is the second of a 3-part sequence of modules that introduces computer programming. It covers Object-Oriented and Functional Programming (including Lazy Evaluation). Language used is Java.
CS5332 Biometric Authentication
Semester 2, AY2020-21
This is a new graduate module about authentication using biometrics. Aimed at senior undergraduates and Master's students, this module will cover: authentication methods; types of biometrics; pattern recognition; standards, user-acceptance, privacy concerns.
A one-off talk offering tips for better technical presentations.
Power Papers 1
Part 1 of How to write technical papers.
Power Papers 2
Part 2 of How to write technical papers.