Peripheral Sweet Spot for Human-like Scanpaths

This project studies how peripheral visual sensing affects human-like scanpath behavior in hard-attention vision models. It investigates the trade-off between recognition accuracy and gaze alignment, showing that human-like scanpaths may emerge most clearly under specific peripheral sensing conditions.

The work uses scanpath and gaze-alignment metrics such as DTW, ScanMatch, NSS, AUC, and GCS, and was accepted as a CogSci 2026 poster with full paper publication.

Keywords: gaze alignment, scanpath similarity, peripheral vision, active vision, cognitive science.