Search papers, labs, and topics across Lattice.
Address:
2
0
2
Directly mapping driver facial cues to traffic objects via cross-attention outperforms intermediate point-of-gaze pipelines, slashing background-object misclassification errors by 49.7%.
Knowing where a driver *isn't* looking is now significantly more accurate, thanks to a scene-aware gaze estimation model that outperforms existing methods by 23.5%.
