Text-Guided Salient Object Detection

Zixian Xu, Luanqi Liu, Yingxun Wang, Xue Wang, Pu Li

2023

Abstract

Salient object detection (SOD), a core task in the field of computer vision, is dedicated to accurately identifying the salient objects in images. Unlike previous research methods, this study recognizes the key role of textual information in salient object detection and thus proposes a unique text-based range control method for salient object detection. In this method, we introduce the semantic labels from the CoSOD3K dataset into a pre-trained text-driven semantic segmentation model to align the textual and image feature information. Subsequently, the image features are analyzed for saliency through a salient object detection network. Through the SFE (Salient Feature Extractor) module, we fuse the extracted salient features with the semantically aligned features to derive the saliency detection results. Experimental results show that the robustness and efficiency of our framework surpass existing salient object detection methods. Users can guide the detection process through natural language interaction, expanding applications such as image editing and data annotation, and to some extent solving challenges like complex backgrounds, multi-scale issues, and blurry boundaries. This offers the potential for new breakthroughs in the field of salient object detection.

Paper Citation

in Harvard Style

Xu Z., Liu L., Wang Y., Wang X. and Li P. (2023). Text-Guided Salient Object Detection. In Proceedings of the 2nd International Seminar on Artificial Intelligence, Networking and Information Technology - Volume 1: ANIT; ISBN 978-989-758-677-4, SciTePress, pages 381-385. DOI: 10.5220/0012284400003807

in Bibtex Style

@conference{anit23,
author={Zixian Xu and Luanqi Liu and Yingxun Wang and Xue Wang and Pu Li},
title={Text-Guided Salient Object Detection},
booktitle={Proceedings of the 2nd International Seminar on Artificial Intelligence, Networking and Information Technology - Volume 1: ANIT},
year={2023},
pages={381-385},
publisher={SciTePress},
organization={INSTICC},
doi={10.5220/0012284400003807},
isbn={978-989-758-677-4},
}

in EndNote Style

TY - CONF

JO - Proceedings of the 2nd International Seminar on Artificial Intelligence, Networking and Information Technology - Volume 1: ANIT
TI - Text-Guided Salient Object Detection
SN - 978-989-758-677-4
AU - Xu Z.
AU - Liu L.
AU - Wang Y.
AU - Wang X.
AU - Li P.
PY - 2023
SP - 381
EP - 385
DO - 10.5220/0012284400003807
PB - SciTePress