YouTubeVIS2019
收藏资源简介:
我们收集了第一个用于视频实例分割的大规模数据集,称为 YouTube-VIS,它基于我们最初的 YouTube-VOS 数据集。具体来说,我们的新数据集具有以下特征。 2019 版本用于 ICCV 2019 的“第 2 届大规模视频对象分割挑战赛”。2,883 个高分辨率 YouTube 视频、2,238 个训练视频、302 个验证视频和 343 个测试视频一个类别标签集,包括 40 个常见对象,例如人、动物和车辆 4,883 个独特的视频实例 131k 高质量手动注释
We present the first large-scale dataset for video instance segmentation, named YouTube-VIS, which is built upon our original YouTube-VOS dataset. Specifically, our new dataset has the following characteristics: The 2019 version was utilized for the "2nd Large-scale Video Object Segmentation Challenge" at ICCV 2019. It comprises 2,883 high-resolution YouTube videos, with 2,238 for training, 302 for validation, and 343 for testing. It employs a single category label set including 40 common object categories such as humans, animals, and vehicles. It includes 4,883 unique video instances and 131k high-quality manual annotations.




