Abstract: Training data for video segmentation are expensive to annotate. This impedes extensions of end-to-end algorithms to new video segmentation tasks, especially in large-vocabulary settings. To ...
Some results have been hidden because they may be inaccessible to you
Show inaccessible results