Gespeichert in:
Bibliographische Detailangaben
Hauptverfasser: Lee, Seung-jae, Seo, Paul Hongsuck
Format: Preprint
Veröffentlicht: 2025
Schlagworte:
Online-Zugang:https://arxiv.org/abs/2506.06537
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
_version_ 1866910993198612480
author Lee, Seung-jae
Seo, Paul Hongsuck
author_facet Lee, Seung-jae
Seo, Paul Hongsuck
contents Audiovisual segmentation (AVS) aims to identify visual regions corresponding to sound sources, playing a vital role in video understanding, surveillance, and human-computer interaction. Traditional AVS methods depend on large-scale pixel-level annotations, which are costly and time-consuming to obtain. To address this, we propose a novel zero-shot AVS framework that eliminates task-specific training by leveraging multiple pretrained models. Our approach integrates audio, vision, and text representations to bridge modality gaps, enabling precise sound source segmentation without AVS-specific annotations. We systematically explore different strategies for connecting pretrained models and evaluate their efficacy across multiple datasets. Experimental results demonstrate that our framework achieves state-of-the-art zero-shot AVS performance, highlighting the effectiveness of multimodal model integration for finegrained audiovisual segmentation.
format Preprint
id arxiv_https___arxiv_org_abs_2506_06537
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Bridging Audio and Vision: Zero-Shot Audiovisual Segmentation by Connecting Pretrained Models
Lee, Seung-jae
Seo, Paul Hongsuck
Computer Vision and Pattern Recognition
Sound
Audio and Speech Processing
Audiovisual segmentation (AVS) aims to identify visual regions corresponding to sound sources, playing a vital role in video understanding, surveillance, and human-computer interaction. Traditional AVS methods depend on large-scale pixel-level annotations, which are costly and time-consuming to obtain. To address this, we propose a novel zero-shot AVS framework that eliminates task-specific training by leveraging multiple pretrained models. Our approach integrates audio, vision, and text representations to bridge modality gaps, enabling precise sound source segmentation without AVS-specific annotations. We systematically explore different strategies for connecting pretrained models and evaluate their efficacy across multiple datasets. Experimental results demonstrate that our framework achieves state-of-the-art zero-shot AVS performance, highlighting the effectiveness of multimodal model integration for finegrained audiovisual segmentation.
title Bridging Audio and Vision: Zero-Shot Audiovisual Segmentation by Connecting Pretrained Models
topic Computer Vision and Pattern Recognition
Sound
Audio and Speech Processing
url https://arxiv.org/abs/2506.06537