UVO Challenge on Video-based Open-World Segmentation 2021: 1st Place Solution

TL;DR

This paper presents a straightforward two-step video instance segmentation method that combines per-frame segmentation with optical flow-based mask matching, achieving first place in the UVO 2021 challenge.

Contribution

The paper introduces a simple yet effective detect-then-match approach for open-world video segmentation, emphasizing high-quality proposals and optical flow for tracking.

Findings

Achieved first place in UVO 2021 challenge

High-quality mask proposals enable simple matching for tracking

Effective use of optical flow improves inter-frame mask matching

Abstract

In this report, we introduce our (pretty straightforard) two-step "detect-then-match" video instance segmentation method. The first step performs instance segmentation for each frame to get a large number of instance mask proposals. The second step is to do inter-frame instance mask matching with the help of optical flow. We demonstrate that with high quality mask proposals, a simple matching mechanism is good enough for tracking. Our approach achieves the first place in the UVO 2021 Video-based Open-World Segmentation Challenge.