Georgia Tech Egocentric Activities - Gaze(+)
Our goal in collecting these datasets is to study the relationship between human activities and their patterns of fixation.
Previous investigations of eye movement have largely been based on studies of static scene viewing, using gaze technology
affixed to a monitor screen. However, to study gaze in the context of object manipulation tasks, a mobile system that captures human gaze in real-life setting is required. We have collected these datasets using recently commercially available eye-tracking systems that record a high-quality video of where people look at and at the same time record their gaze location in every frame of the video.
Please refer to the following paper when using the datasets, code, software:
- Alireza Fathi, Yin Li, James M. Rehg, Learning to recognize daily actions using gaze, ECCV 2012 (PDF).