Search papers, labs, and topics across Lattice.
3
0
5
ARTEMIS transforms weakly supervised video polyp segmentation by evolving reliable masks over time, achieving unprecedented accuracy in challenging clinical settings.
Even Gemini struggles with cinematic expressiveness, revealing that current multi-talker audio-video generation models still have a long way to go in capturing the nuances of acting, narrative, and atmosphere.
Explicitly modeling physiological frequency priors in rPPG unlocks significant improvements in signal recovery, even under challenging motion conditions that plague existing time-domain methods.
