The 2019 DAVIS Challenge on VOS: Unsupervised Multi-Object Segmentation
Sergi Caelles, Jordi Pont-Tuset, Federico Perazzi, Alberto Montes, Kevis-Kokitsi Maninis, Luc Van Gool
TL;DR
The paper expands DAVIS to include an unsupervised multi-object video object segmentation track, addressing the absence of test-time supervision and the need for consistent multi-object segmentation. It defines the task precisely, re-annotates DAVIS 2017 train/val for semantic consistency, and introduces new test-dev/test-challenge sets, using a bipartite matching framework over the J&F metric with a pool of video object proposals; RVOS is evaluated as a zero-shot baseline. The results demonstrate the substantial gap between unsupervised and semi-supervised VOS, highlighting the challenging nature of discovering and consistently tracking multiple objects without annotations. Overall, the work provides a standardized benchmark, clear evaluation protocol, and practical baseline to drive progress toward fully automatic, multi-object VOS research.
Abstract
We present the 2019 DAVIS Challenge on Video Object Segmentation, the third edition of the DAVIS Challenge series, a public competition designed for the task of Video Object Segmentation (VOS). In addition to the original semi-supervised track and the interactive track introduced in the previous edition, a new unsupervised multi-object track will be featured this year. In the newly introduced track, participants are asked to provide non-overlapping object proposals on each image, along with an identifier linking them between frames (i.e. video object proposals), without any test-time human supervision (no scribbles or masks provided on the test video). In order to do so, we have re-annotated the train and val sets of DAVIS 2017 in a concise way that facilitates the unsupervised track, and created new test-dev and test-challenge sets for the competition. Definitions, rules, and evaluation metrics for the unsupervised track are described in detail in this paper.