Improving Unsupervised Video Object Segmentation via Fake Flow Generation

TL;DR

This paper introduces a novel data generation method that creates fake optical flow data from single images to enhance unsupervised video object segmentation, achieving state-of-the-art results without complex modules.

Contribution

A new approach to generate fake optical flows from single images improves training data availability for VOS, leading to better performance.

Findings

Achieved state-of-the-art results on benchmark datasets.

Generated large-scale training data from single images.

Improved VOS performance without complex modules.

Abstract

Unsupervised video object segmentation (VOS), also known as video salient object detection, aims to detect the most prominent object in a video at the pixel level. Recently, two-stream approaches that leverage both RGB images and optical flow maps have gained significant attention. However, the limited amount of training data remains a substantial challenge. In this study, we propose a novel data generation method that simulates fake optical flows from single images, thereby creating large-scale training data for stable network learning. Inspired by the observation that optical flow maps are highly dependent on depth maps, we generate fake optical flows by refining and augmenting the estimated depth maps of each image. By incorporating our simulated image-flow pairs, we achieve new state-of-the-art performance on all public benchmark datasets without relying on complex modules. We believe that our data generation method represents a potential breakthrough for future VOS research.