Reproduce “FSDAM: Few-Shot Driving Attention Modeling via Vision-Language Coupling” 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
Kaiser Hamid, Can Cui, Khandakar Ashrafi Akbar, Ziran Wang, Nade Liang
FSDAM predicts driver gaze maps and structured explanations from a single image using separate gaze and caption pathways linked by training-only vision-language alignment. The model uses 90 task-specific annotated examples; future-focus labels come from paired frames. On BDD-A, it reaches CC 0.60 and KL 1.13, versus LLada’s 0.37 and 1.86 on the same 90 examples, but does not establish safety or faithful explanations.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.