Reproduce “MIDAS: Deep learning human action intention prediction from natural eye movement patterns” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Paul Festor, Ali Shafti, Alex Harston, Michey Li, Pavel Orlov, A. Aldo Faisal
MIDAS distinguishes inspection from intended manipulation in instructed table tasks by analyzing sequences of natural gaze alongside visual context about target objects. Its bidirectional long short-term memory layers process gaze and object-position time series. In five-fold cross-validation, it achieved 91.9% mean accuracy, versus 84.5% for gradient boosting; this result concerns the study’s two-label task, not a deployed general-purpose interface.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.