Visitor and Kangaroo: A Yarbus Replication Dataset

The Visitor and Kangaroo dataset is a large-scale, laboratory-based eye-tracking benchmark designed for sequential and AOI-based gaze analysis. The Visitor dataset includes recordings from 144 participants who viewed the same painting under seven task conditions, closely replicating the classical Yarbus paradigm in a controlled setting. Semantically defined Areas of Interest (AOIs) enable systematic investigation of gaze transitions, revisit dynamics, and scanpath sequences across tasks and individuals. The complementary Kangaroo dataset contains eight logic problem stimuli inspired by the Math Kangaroo competition, supporting independent research on visual problem-solving strategies and reasoning processes. The dataset is released in open formats and includes raw gaze data, event-level exports, AOI-based metrics, calibration data, AOI definitions, and stimulus materials. It provides a reproducible and scalable resource for methodological development, scanpath comparison, sequence modeling, task decoding, and research on task-driven attention.