Universiteit Leiden

nl en

LIACS alumnus receives ACM Multimedia Rising Star Award

Former PhD student of the Leiden Institute of Advanced Computer Science (LIACS) Bart Thomee has received this year’s Rising Star Award from the Multimedia group of ACM, the worldwide computer science association. Today, Thomee is a successful researcher at Yahoo Labs.

Understand search intent faster

Dr. Bart Thomee finished his doctorate research at the Leiden Institute of Advanced Computer Science (LIACS) in 2010. In Leiden, he published some of the ground breaking work on artificial imagination within the LIACS Media Lab (Media Research and Data Mining Group). He managed to understand people’s search intent up to 23% more rapidly by generating imagery that resemble the ideal image they are looking for. Since then, numerous research groups have started using artificially synthesized imagery for improving all aspects of computer vision.

Duplicate detection

Thomee also worked on very large scale near-duplicate visual detection. He designed new descriptors that could compactly represent web-scale image collections and accurately detect transformed versions of them. He was given an ACM Best Paper Citation for this research, which was part of the Computer Systems, Imagery and Media Programme from the Media Research group. Today, this group is employing deep learning approaches toward learning optimal features from big data.

Geographic computing for Yahoo

At Yahoo Labs, Thomee started to work on geographic computing. He developed a technique that could process billions of georeferenced labels in a matter of hours. His work became a reference example at Yahoo. Furthermore, Thomee began using compass information from camera phone images. He made it possible to accurately pinpoint the locations and surface area of landmarks, only based on pictures that may have been taken up to several kilometers away.

Open dataset

With this Rising Star Award, Thomee is also honoured for his efforts to open the Yahoo Flickr Creative Commons 100 Million (YFCC100M) dataset to scientists worldwide, but also to the public. YFCC100M is the largest public multimedia collection ever released, with a total of 100 million media objects. Since February 2016, it has been requested over 1200 times.