Boosting Unsupervised Video Instance Segmentation with Automatic Quality-Guided Self-Training

Fuente: arXiv
Enregistré dans:
Détails bibliographiques
Auteurs principaux: Lu, Kaixuan, Kaya, Mehmet Onurcan, Papadopoulos, Dim P.
Format: Preprint
Publié: 2025
Sujets:
Accès en ligne:
Tags: Ajouter un tag
Pas de tags, Soyez le premier à ajouter un tag!
_version_ 1866908697428492288
author Lu, Kaixuan
Kaya, Mehmet Onurcan
Papadopoulos, Dim P.
author_facet Lu, Kaixuan
Kaya, Mehmet Onurcan
Papadopoulos, Dim P.
contents Video Instance Segmentation (VIS) faces significant annotation challenges due to its dual requirements of pixel-level masks and temporal consistency labels. While recent unsupervised methods like VideoCutLER eliminate optical flow dependencies through synthetic data, they remain constrained by the synthetic-to-real domain gap. We present AutoQ-VIS, a novel unsupervised framework that bridges this gap through quality-guided self-training. Our approach establishes a closed-loop system between pseudo-label generation and automatic quality assessment, enabling progressive adaptation from synthetic to real videos. Experiments demonstrate state-of-the-art performance with 52.6 $\text{AP}_{50}$ on YouTubeVIS-2019 $\texttt{val}$ set, surpassing the previous state-of-the-art VideoCutLER by 4.4%, while requiring no human annotations. This demonstrates the viability of quality-aware self-training for unsupervised VIS. We will release the code at https://github.com/wcbup/AutoQ-VIS.
format Preprint
id arxiv_https___arxiv_org_abs_2512_06864
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Boosting Unsupervised Video Instance Segmentation with Automatic Quality-Guided Self-Training
Lu, Kaixuan
Kaya, Mehmet Onurcan
Papadopoulos, Dim P.
Computer Vision and Pattern Recognition
Video Instance Segmentation (VIS) faces significant annotation challenges due to its dual requirements of pixel-level masks and temporal consistency labels. While recent unsupervised methods like VideoCutLER eliminate optical flow dependencies through synthetic data, they remain constrained by the synthetic-to-real domain gap. We present AutoQ-VIS, a novel unsupervised framework that bridges this gap through quality-guided self-training. Our approach establishes a closed-loop system between pseudo-label generation and automatic quality assessment, enabling progressive adaptation from synthetic to real videos. Experiments demonstrate state-of-the-art performance with 52.6 $\text{AP}_{50}$ on YouTubeVIS-2019 $\texttt{val}$ set, surpassing the previous state-of-the-art VideoCutLER by 4.4%, while requiring no human annotations. This demonstrates the viability of quality-aware self-training for unsupervised VIS. We will release the code at https://github.com/wcbup/AutoQ-VIS.
title Boosting Unsupervised Video Instance Segmentation with Automatic Quality-Guided Self-Training
topic Computer Vision and Pattern Recognition
url https://arxiv.org/abs/2512.06864