I-MPN: inductive message passing network for efficient human-in-the-loop annotation of mobile eye tracking data

Area of Science:

• Cognitive Psychology
• Computer Vision
• Human-Computer Interaction

Background:

• Understanding human visual processing in dynamic environments is key for psychology and user-centered design.
• Manual analysis of mobile eye-tracking data is time-consuming and inefficient.
• Automated methods are needed to analyze complex visual information from egocentric recordings.

Purpose of the Study:

• To develop a novel human-centered learning algorithm for automated object recognition in mobile eye-tracking settings.
• To improve the efficiency and accuracy of analyzing visual data from dynamic environments.
• To enable better understanding of human visual information processing.

Main Methods:

• Integration of an object detector with a spatial relation-aware inductive message-passing network (I-MPN).
• Utilizing node profile information and object correlations for learning embedding functions.
• Employing an interactive-based learning approach with user feedback for smaller annotated samples.

Main Results:

• Significant performance improvements over fixed training/testing algorithms on three distinct video sequences.
• Demonstrated efficiency in data annotation processes.
• Outperformed prior interactive methods in object recognition tasks within mobile eye-tracking data.

Conclusions:

• The proposed algorithm offers an efficient and effective solution for automated object recognition in mobile eye-tracking.
• The human-centered learning approach facilitates rapid adaptation and efficient reasoning in dynamic contexts.
• This method advances the analysis of visual information processing in real-world environments.