YouTube-VOS: Sequence-to-Sequence Video Object Segmentation

TL;DR

This paper introduces a large-scale video object segmentation dataset, YouTube-VOS, and proposes a sequence-to-sequence network that effectively captures long-term spatial-temporal features, achieving state-of-the-art results.

Contribution

The paper presents the largest video object segmentation dataset to date and a novel sequence-to-sequence model for improved long-term video segmentation.

Findings

Achieved top performance on YouTube-VOS test set

Comparable results on DAVIS 2016 with current state-of-the-art

Large dataset significantly improves model effectiveness

Abstract

Learning long-term spatial-temporal features are critical for many video analysis tasks. However, existing video segmentation methods predominantly rely on static image segmentation techniques, and methods capturing temporal dependency for segmentation have to depend on pretrained optical flow models, leading to suboptimal solutions for the problem. End-to-end sequential learning to explore spatial-temporal features for video segmentation is largely limited by the scale of available video segmentation datasets, i.e., even the largest video segmentation dataset only contains 90 short video clips. To solve this problem, we build a new large-scale video object segmentation dataset called YouTube Video Object Segmentation dataset (YouTube-VOS). Our dataset contains 3,252 YouTube video clips and 78 categories including common objects and human activities. This is by far the largest video object segmentation dataset to our knowledge and we have released it at https://youtube-vos.org. Based on this dataset, we propose a novel sequence-to-sequence network to fully exploit long-term spatial-temporal information in videos for segmentation. We demonstrate that our method is able to achieve the best results on our YouTube-VOS test set and comparable results on DAVIS 2016 compared to the current state-of-the-art methods. Experiments show that the large scale dataset is indeed a key factor to the success of our model.