Register now After registration you will be able to apply for this opportunity online.
This opportunity is not published. No applications will be accepted.
Gaze-based Interaction with Public Display by Personal Eye Image Generation
In this project, we aim to develop an interactive system with a public display that can perform immediate personal gaze-based interaction.
Keywords: Gaze-based interaction; Generative model
Gaze-based interaction with public display is a specific research topic that has not been solved yet. The difficulty of the setting is two-fold: (1) The commercial eye trackers cannot be used due to the long distance between user and camera, therefore only webcam can be deployed; (2) There is no much time to collect personal training data so that gaze estimation accuracy is not sufficient enough for fine-level interaction. In the computer vision community, there is already works that can generate the personal eye images with desired gaze directions and head poses with just a few calibration samples. However, these works have not been proved to work in real-world applications.
In this project, we aim to develop an interactive system with a public display that can perform immediate personal gaze-based interaction. We will implement the current personal eye samples generation method with GAN model, and evaluate the gaze estimation performance with few or single calibration samples. We then need to design the interactive interface to retrieve the single/few calibration samples and evaluate the system with users in real-world settings.
Gaze-based interaction with public display is a specific research topic that has not been solved yet. The difficulty of the setting is two-fold: (1) The commercial eye trackers cannot be used due to the long distance between user and camera, therefore only webcam can be deployed; (2) There is no much time to collect personal training data so that gaze estimation accuracy is not sufficient enough for fine-level interaction. In the computer vision community, there is already works that can generate the personal eye images with desired gaze directions and head poses with just a few calibration samples. However, these works have not been proved to work in real-world applications. In this project, we aim to develop an interactive system with a public display that can perform immediate personal gaze-based interaction. We will implement the current personal eye samples generation method with GAN model, and evaluate the gaze estimation performance with few or single calibration samples. We then need to design the interactive interface to retrieve the single/few calibration samples and evaluate the system with users in real-world settings.
The proposed system will be able to generate the personal eye images on the fly and perform accurate personal gaze estimation, which then is used for gaze-based interactive system in real-world settings.
The proposed system will be able to generate the personal eye images on the fly and perform accurate personal gaze estimation, which then is used for gaze-based interactive system in real-world settings.
Dr. Xucong Zhang, Postdoctoral researcher, AIT Lab, ETH Zurich. Email: xucong.zhang@inf.ethz.ch webpage: https://ait.ethz.ch/people/zhang/
Dr. Xucong Zhang, Postdoctoral researcher, AIT Lab, ETH Zurich. Email: xucong.zhang@inf.ethz.ch webpage: https://ait.ethz.ch/people/zhang/