Computer Vision

Computer Vision is a classic research area since the 1950s, when the earliest computers struggled to interpret photography and scenes from the natural world. Much like in other fields of artificial intelligence, the last decade has seen deep learning and neural networks revolutionize computer vision. The ability of cameras to tease apart complex scenes, peer through stormy weather, or detect anomalies in security applications rely on continued scientific discovery and technology development by the world’s brightest.

A*STAR’s computer vision efforts reside in a sweet spot between theoretical computer science research and application-driven AI productization, and are carried out at several Research Institutes in multi-disciplinary teams. Our scientists continue to push the boundaries of computer vision while inspired by the practical demands of our fast-paced techcentric society.

Computer Vision

Key Researchers

Lim Joo Hwee, A*STAR Institute for Infocomm Research (A*STAR I2R)
, A*STAR Institute for Infocomm Research (A*STAR I2R)
, A*STAR Institute for Infocomm Research (A*STAR I2R)
, A*STAR Institute for Infocomm Research (A*STAR I2R)
Dr Lee Hwee Kuan, A*STAR Bioinformatics Institute (A*STAR BII)
Dr Yu Weimiao, A*STAR Bioinformatics Institute (A*STAR BII)

Key Projects

Co-development of Senior Care Technology with AI Thinktank òòò½Íø(MI, A*STAR I2R)

CareCam is a Singapore-based A*STAR spinoff that develops machine vision technology for the healthcare industry. Their flagship product is CareCam 3D-Gait, an iOS mobile app that uses computer vision and AI to assess and measure the risk of medical disorders by analyzing gait. CareCam 3D-Gait is used to identify mobility deficiencies quickly and provide quantitative gait metrics against normative data in real-time. The app leverages on the integration of AI models for detection, tracking, phase segmentation, pose estimation and kinematics analysis for on-board processing and assessment. This information can be used by healthcare professionals to improve patient care and prevent injuries.

Improving 3D recognition performance with minimum extra costs for vision guided robotics (MI, I2R)

3D recognition is a key technique that has wide applications in vision guided robotics (VGR), e.g., autonomous vehicles, navigation, surveillance and beyond. This project aims to improve the accuracy and robustness of 3D recognition models with minimum extra costs. For accuracy, we propose to exploit different sorts of unlabelled data which are cheap to obtain. For robustness, we focus on the open-set generalization challenge and propose to learn few-shot learning models to endow models the ability of generalizing to novel objects with only a few labelled examples.
vgr

A*STAR celebrates International Women's Day

A*STAR Women In Science

From groundbreaking discoveries to cutting-edge research, our researchers are empowering the next generation of female science, technology, engineering and mathematics (STEM) leaders.