Target-aware video object segmentation using prompt guidance

Bo Miao

doi:10.26182/qthp-5d95

Target-aware video object segmentation using prompt guidance

Bo Miao

School of Physics, Mathematics and Computing

Research output: Thesis › Doctoral Thesis

194 Downloads (Pure)

Abstract

Video object segmentation aims to separate user-specified objects from backgrounds in unconstrained videos, fuelling diverse applications such as robotics, sports analytics, augmented reality, and video editing. Although prior works explored various prompts such as masks, scribbles, or text descriptions for specifying target objects, they face significant challenges in accuracy (temporal consistency, vision-language alignment) and applicability (versatility, efficiency, and annotation costs). This dissertation systematically addresses these core challenges by introducing novel approaches and benchmarks to advance this field.

Original language	English
Qualification	Doctor of Philosophy
Awarding Institution	The University of Western Australia
Supervisors/Advisors	Mian, Ajmal, Supervisor Bennamoun, Mohammed, Supervisor
Award date	16 Apr 2025
DOIs	https://doi.org/10.26182/qthp-5d95
Publication status	Unpublished - 2025

Funding

Funders	Funder number
ARC Australian Research Council	IH180100002

Access to Document

10.26182/qthp-5d95

THESIS - DOCTOR OF PHILOSOPHY - MIAO Bo - 2025
This work is protected by Copyright. You may print or download ONE copy of this document for the purpose of your own non-commercial research or study. Any other use requires permission from the copyright owner. The Copyright Act requires you to attribute any copyright works you quote or paraphrase.

ARC Research Hub for Driving Farming Productivity and Disease Prevention
Bennamoun, M. (Investigator 01) & Mian, A. (Investigator 02)
ARC Australian Research Council
1/01/19 → 31/12/23
Project: Research

3 Conference paper
2 Article
1 Preprint

Referring Human Pose and Mask Estimation in the Wild.
Miao, B., Feng, M., Wu, Z., Bennamoun, M., Gao, Y. & Mian, A., 2024, Neural Information Processing Systems Foundation, Inc.
Research output: Working paper › Preprint
Region Aware Video Object Segmentation With Deep Motion Modeling
Miao, B., Bennamoun, M., Gao, Y. & Mian, A., 29 Mar 2024, In: IEEE Transactions on Image Processing. 33, p. 2639-2651 13 p.
Research output: Contribution to journal › Article › peer-review
18 Link opens in a new tab Citations (Scopus)
Temporally Consistent Referring Video Object Segmentation with Hybrid Memory
Miao, B., Bennamoun, M., Gao, Y., Shah, M. & Mian, A., 2024, In: IEEE Transactions on Circuits and Systems for Video Technology. 34, 11, p. 11373-11385 13 p., 10572009.
Research output: Contribution to journal › Article › peer-review

Open Access
File
21 Link opens in a new tab Citations (Scopus)

169 Downloads (Pure)

Cite this

@phdthesis{c8d846d834774b4d9da38dba407c5db7,

title = "Target-aware video object segmentation using prompt guidance",

abstract = "Video object segmentation aims to separate user-specified objects from backgrounds in unconstrained videos, fuelling diverse applications such as robotics, sports analytics, augmented reality, and video editing. Although prior works explored various prompts such as masks, scribbles, or text descriptions for specifying target objects, they face significant challenges in accuracy (temporal consistency, vision-language alignment) and applicability (versatility, efficiency, and annotation costs). This dissertation systematically addresses these core challenges by introducing novel approaches and benchmarks to advance this field.",

keywords = "video object segmentation, video understanding, vision-and-language, human-AI interaction, segmentation, pose estimation, temporal consistency",

author = "Bo Miao",

year = "2025",

doi = "10.26182/qthp-5d95",

language = "English",

school = "The University of Western Australia",

}

TY - BOOK

T1 - Target-aware video object segmentation using prompt guidance

AU - Miao, Bo

PY - 2025

Y1 - 2025

N2 - Video object segmentation aims to separate user-specified objects from backgrounds in unconstrained videos, fuelling diverse applications such as robotics, sports analytics, augmented reality, and video editing. Although prior works explored various prompts such as masks, scribbles, or text descriptions for specifying target objects, they face significant challenges in accuracy (temporal consistency, vision-language alignment) and applicability (versatility, efficiency, and annotation costs). This dissertation systematically addresses these core challenges by introducing novel approaches and benchmarks to advance this field.

AB - Video object segmentation aims to separate user-specified objects from backgrounds in unconstrained videos, fuelling diverse applications such as robotics, sports analytics, augmented reality, and video editing. Although prior works explored various prompts such as masks, scribbles, or text descriptions for specifying target objects, they face significant challenges in accuracy (temporal consistency, vision-language alignment) and applicability (versatility, efficiency, and annotation costs). This dissertation systematically addresses these core challenges by introducing novel approaches and benchmarks to advance this field.

KW - video object segmentation

KW - video understanding

KW - vision-and-language

KW - human-AI interaction

KW - segmentation

KW - pose estimation

KW - temporal consistency

U2 - 10.26182/qthp-5d95

DO - 10.26182/qthp-5d95

M3 - Doctoral Thesis

ER -

Target-aware video object segmentation using prompt guidance

Abstract

Funding

Access to Document

Fingerprint

Projects

ARC Research Hub for Driving Farming Productivity and Disease Prevention

Research output

Referring Human Pose and Mask Estimation in the Wild.

Region Aware Video Object Segmentation With Deep Motion Modeling

Temporally Consistent Referring Video Object Segmentation with Hybrid Memory

Cite this