26 September 2017 Department of Computer Science Faculty

 

26 September 2017 – Associate Professor Terence Sim won the Test of Time Award at the 12th IEEE International Conference on Automatic Face and Gesture Recognition (FG 2017).

The conference was held in Washington, D.C. from 30 May to 3 June this year.

The award acknowledges past FG conference papers, published in the last 15 to 20 years, that have received the most citations and are recognised to be influential to current researchers today.

According to Google Scholar, Prof Sim’s 2002 paper, The CMU Pose, Illumination, and Expression (PIE) Database, has been cited by over 2900 other papers.

“In other words, our paper stood the test of time,” said Prof Sim.

Prof Sim wrote the paper during his stint as a PhD student in Carnegie Mellon University (CMU). He co-authored the paper with Simon Baker and Maan Bsat from CMU. The paper was published in the 2002 Proceedings of Fifth IEEE International Conference on Automatic Face Gesture Recognition.

Prof Sim and his collaborators contributed a key database of facial images, captured under different Pose, Illumination and facial Expression (PIE). They collected over 40,000 facial images from over 70 volunteers in various illumination conditions and different expressions.

“Our database was created because there were few facial images databases in 2002 and they did not have the image diversity required to advance facial recognition,” explained Prof Sim.

Hence the team methodically captured more difficult images, under controlled situations, in both images and videos.

“At the point of publication, our database was the largest, at over 40GB, and most diverse, in terms of types of images captured. It was very valuable for researchers to benchmark and test new algorithms,” said Prof Sim.

FG 2017 is an international forum for research in image and video-based face, gesture and body movement recognition. Its scope includes advances in fundamental computer vision, pattern recognition, computer graphics, machine learning techniques relevant to face, gesture, and body motion, and algorithms and analysis of specific applications.