Ref-Youtube-VOS

该数据集是一个大规模的参照视频对象分割数据集，包含了约15,000个参照表达式，对应3,900多个视频，每个视频的每第五帧都提供了像素级的实例分割标注。值得注意的是，验证集的标注并不公开。该数据集的规模属于大型，主要针对的任务是参照视频对象分割。

This dataset is a large-scale referring video object segmentation dataset, which contains approximately 15,000 referring expressions corresponding to over 3,900 videos. Pixel-level instance segmentation annotations are provided for every fifth frame of each video. Notably, the annotations of the validation set are not publicly available. Given its large scale, this dataset is primarily targeted at the task of referring video object segmentation.

搜集汇总

背景与挑战

背景概述

Ref-Youtube-VOS是一个大型参照视频对象分割数据集，包含约15,000个参照表达式和超过3,900个视频，每第五帧提供像素级实例分割标注，但验证集标注不公开，专为参照视频对象分割任务设计。