1 paper
Sangbeom Lim, Seongchan Kim, Seungjun An +3
Current benchmarks for video segmentation are limited to annotating only salient objects (i.e., foreground instances). Despite their impressive architectural designs, previous work…