Best Student Paper Honors

A paper co-authored by graduate students advised by MD2K Deputy Director James M. Rehg earned a best student paper prize at the 11th IEEE International Conference on Automatic Face and Gesture Recognition (FG 2015), held May 4-8 in Ljubljana, Slovenia.

In addition, two other papers co-authored by Rehg’s students will be presented at a top conference on computer vision in June.

None of the papers was directly supported by MD2K; each involved work done before MD2K received its award from the National Institutes of Health in October as part of the Big Data to Knowledge Initiative. However, the work is noteworthy because the papers all involve first-person vision and describe methods that will likely be used in future MD2K research.

The paper that earned the honors, Detecting Bids for Eye Contact Using a Wearable Camera was authored by Zhefan Ye, Yin Li, Yun Liu, Chanel Bridges, Agata Rozga, James M. Rehg. It presented findings from a study that explored using a point-of-view camera worn by an adult to capture first-person views of the child-adult interaction from the adult’s perspective. The videos were analyzed to automatically identify the moments during which the child was trying to make eye contact with the adult. Researchers used that data to develop a method that was tested on a dataset of 12 children. Their method outperformed state-of-the-art approaches and enables measuring gaze behavior in real-world social interactions.

Ye, Li and Liu are graduate students advised by Rehg. For more information about the project, go here.

The other two papers will be presented at the IEEE Conference on Computer Vision and Pattern Recognition (CVPR), to be held June 8-12 in Boston, Mass. CVPR is the top conference on computer vision and also the top-ranking conference in computer science and engineering, based on Google’s h5 index.

Delving into Egocentric Actions, authored by Yin Li, Zhefan Ye, and James M. Rehg, addresses the problem of recognizing actions based on video captured from a head-worn camera (egocentric video). It demonstrates that novel egocentric features containing information about head movement, location of hands, and object detections are superior to traditional video features for action recognition.

Gaze-Engabled Egocentric Video Summarization via Constrained Submodular Maximization, authored by Jia Xu, Lopamudra Mukherjee, Yin Li, Jamieson Warner, James M. Rehg, and Vikas Singh, presents a method for summarizing egocentric videos captured from a wearable camera system that incorporates eye-tracking. Researchers demonstrated that gaze information obtained from eye tracking can be used to construct superior summarizations in comparison to standard video summarization methods. This project is a collaboration between two BD2K centers: the research groups of Jim Rehg in MD2K and Vikas Singh in the Center for Predictive Computational Phenotyping (CPCP) led by U. Wisconsin.

MD2K is creating tools that will make it easier to develop, analyze and interpret health data generated by mobile and wearable sensors, with a goal of developing big data solutions that will reliably quantify the factors, both environmental and social, that contribute to health and disease risk.

Here are citations all three papers:

Zhefan Ye, Yin Li, Yun Liu, Chanel Bridges, Agata Rozga, James M. Rehg. Detecting Bids for Eye Contact Using a Wearable Camera. In Proc. 11th IEEE International Conference on Automatic Face and Gesture Recognition (FG 2015), Ljubljana, Slovenia, May 4-8, 2015. Project website. Award.

Yin Li, Zhefan Ye, and James M. Rehg. Delving Into Egocentric Actions. In Proc. IEEE Conf. on Computer Vision and Pattern Recognition (CVPR), Boston, MA, June, 2015.

Jia Xu, Lopamudra Mukherjee, Yin Li, Jamieson Warner, James M. Rehg, and Vikas Singh. Gaze-Enabled Egocentric Video Summarization via Constrained Submodular Maximization. In Proc. IEEE Conf. on Computer Vision and Pattern Recognition (CVPR), Boston, MA, June, 2015.

Copyright © 2017 MD2K. MD2K is supported by the National Institutes of Health Big Data to Knowledge Initiative (Grant #1U54EB020404)

Team: Cornell Tech, GA Tech, Harvard, U. Memphis, Northwestern, Ohio State, Open mHealth, UCLA, UCSD, UCSF, UMass, U. Michigan, Utah, WVU