RVOS: End-to-End Recurrent Network for Video Object Segmentation

TL;DR

This paper introduces RVOS, an end-to-end recurrent neural network for zero-shot and one-shot video object segmentation, demonstrating competitive accuracy and faster inference on DAVIS-2017 and YouTube-VOS benchmarks.

Contribution

The paper presents the first fully end-to-end recurrent network for zero-shot video object segmentation, with novel recurrence on spatial and temporal domains.

Findings

Achieves state-of-the-art results on DAVIS-2017 without online learning.

Performs comparably to top methods on YouTube-VOS.

Runs at 44ms/frame, faster than previous approaches.

Abstract

Multiple object video object segmentation is a challenging task, specially for the zero-shot case, when no object mask is given at the initial frame and the model has to find the objects to be segmented along the sequence. In our work, we propose a Recurrent network for multiple object Video Object Segmentation (RVOS) that is fully end-to-end trainable. Our model incorporates recurrence on two different domains: (i) the spatial, which allows to discover the different object instances within a frame, and (ii) the temporal, which allows to keep the coherence of the segmented objects along time. We train RVOS for zero-shot video object segmentation and are the first ones to report quantitative results for DAVIS-2017 and YouTube-VOS benchmarks. Further, we adapt RVOS for one-shot video object segmentation by using the masks obtained in previous time steps as inputs to be processed by the recurrent module. Our model reaches comparable results to state-of-the-art techniques in YouTube-VOS benchmark and outperforms all previous video object segmentation methods not using online learning in the DAVIS-2017 benchmark. Moreover, our model achieves faster inference runtimes than previous methods, reaching 44ms/frame on a P100 GPU.