Sparse Self-Prompt-Guided Stereo Matching for Real-World Generalization
Sensors · 17.05.2026 · 26 (10) · 3173
DOI: 10.3390/s26103173 ↗Kurz gefasst
Fachartikel in der Zeitschrift Sensors, online erschienen am 17.05.2026. Verfasst von Hangbiao Li und 6 weiteren Autor:innen. Beteiligte Einrichtung: School of Information and Engineering, Nanchang Hangkong University, Nanchang 330063, China. Die Arbeit nennt 50 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0). Gefördert von: National Natural Science Foundation of China, Jiangxi Natural Science Foundation, Jiangxi Natural Science Foundation, China.
Abstract (Originalsprache)
Stereo matching has witnessed rapid advances on curated benchmarks, yet deploying models in unconstrained real-world environments remains a fundamental challenge. This paper presents a sparse self-prompt-guided network (SSPGNet) for stereo matching with strong generalization across diverse environments. Our core innovation lies in a sparse self-prompt guidance mechanism: (1) a sparse disparity map, used as a prompt, is self-estimated from visual foundation model features via cost aggregation; (2) the sparse disparity is progressively refined into dense disparity maps through cross-attention-based stereo feature interaction, enabling sparse-to-dense disparity prediction. Additionally, we collected a diverse set of indoor and outdoor stereo pairs by using a ZED 2 camera to assess the real-world performance of our model. Extensive experiments demonstrate that the proposed sparse-to-dense prompt mechanism not only preserves the semantic awareness of visual foundation models but also enhances stereo correspondence reasoning, achieving strong performance on public benchmarks and our in-the-wild dataset. Specifically, under the cross-domain (zero-shot) protocol, the proposed SSPGNet achieves bad-pixel error rates of 3.6% on KITTI 2012 (>3 px), 4.4% on KITTI 2015 (>3 px), 7.6% on Middlebury (>2 px), and 2.1% on ETH3D (>1 px), ranking first on three of the four public benchmarks. These results highlight the potential of SSPGNet for direct deployment in real-world stereo perception systems. The code is publicly available at GitHub.
Schlagwörter
Stichwörter der Autor:innen
Autor:innen und Institutionen
- School of Information and Engineering, Nanchang Hangkong University, Nanchang 330063, China.
- North Lian Chuang Communication Co., Ltd., Nanchang 330096, China.
- School of Computer Science and Technology, University of Science and Technology of China, Hefei 230026, China.
- School of Instrumentation and Optoelectronic Engineering, Beihang University, Beijing 100191, China.
ORCID-Profile: Zhibo Rao
Förderung
- National Natural Science Foundation of China
- Jiangxi Natural Science Foundation
- Jiangxi Natural Science Foundation, China
Zitieren
APA
Li, H., Mo, H., Li, X., Fang, T., Liu, S., Yu, S., & Rao, Z. (2026). Sparse Self-Prompt-Guided Stereo Matching for Real-World Generalization. Sensors, 26(10), 3173. https://doi.org/10.3390/s26103173
Vancouver
Li H, Mo H, Li X, Fang T, Liu S, Yu S, et al. Sparse Self-Prompt-Guided Stereo Matching for Real-World Generalization. Sensors. 2026;26(10):3173. doi:10.3390/s26103173
BibTeX
@article{Li2026,
author = {Li, Hangbiao and Mo, Haojun and Li, Xing and Fang, Tao and Liu, Sikun and Yu, Shuzhen and Rao, Zhibo},
title = {Sparse Self-Prompt-Guided Stereo Matching for Real-World Generalization},
journal = {Sensors},
year = {2026},
volume = {26},
number = {10},
pages = {3173},
doi = {10.3390/s26103173}
}
EUVIMED