Abstract: This paper focuses on the mask utilization of video object segmentation (VOS). The mask here mains the reference masks in the memory bank, i.e., several chosen high-quality predicted masks, ...
Some results have been hidden because they may be inaccessible to you
Show inaccessible results