VisualRNet: Lightweight Camera Rotation Estimation from Low-Resolution Optical Flow via Cross-Modal Supervision
Sensors · 24.04.2026 · 26 (9) · 2655
DOI: 10.3390/s26092655 ↗Kurz gefasst
Fachartikel in der Zeitschrift Sensors, online erschienen am 24.04.2026. Verfasst von Xiong Yang und 2 weiteren Autor:innen. Beteiligte Einrichtung: School of Cyberspace Security, Changzhou College of Information Technology, Changzhou 213164, China. Die Arbeit nennt 32 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0). Gefördert von: the Artificial Intelligence Empowered Curriculum and 471 Textbook System Construction Project for Higher Vocational Education of the Ministry of Education,, Artificial Intelligence Empowered Curriculum and Textbook System Construction Project for Higher Vocational Education of the Ministry of Education, Industry-University-Research Collaboration Project of the Science and Technology Department of Jiangsu Province.
Abstract (Originalsprache)
Camera rotation estimation is a key component in video stabilization and motion analysis. In many practical scenarios, inertial measurements are unavailable or temporally unreliable, while classical geometric pipelines degrade under blur, low texture, and low illumination. This paper investigates whether substantially downsampled optical flow can retain sufficient structure for accurate frame-to-frame rotation regression. We present VisualRNet, a lightweight rotation-specific visual regression framework trained with cross-modal IMU supervision. Our design uses coordinate-aware feature encoding, depthwise separable convolutions, lightweight attention, and a compact 6D rotation head to model the spatial structure of rotational flow fields. On Deep-FVS, VisualRNet achieves a mean rotation error of 0.3151° on the test set. The VisualRNet regression head contains 7.7 K parameters, 0.002 GFLOPs, and runs at 729 FPS, while the full pipeline with the FastFlowNetv2 frontend contains 1.374 M parameters, 7.194 GFLOPs, and runs at approximately 113 FPS. A cross-camera adaptation experiment on TUM VI further indicates that the learned motion representation can be aligned to a new camera system with limited calibration data. These results support low-resolution optical flow as a practical input for visual rotation estimation and suggest particular value in stabilization-oriented and cost-sensitive applications where approximate rotational trend matters more than full scene geometry.
Schlagwörter
Stichwörter der Autor:innen
Autor:innen und Institutionen
- School of Cyberspace Security, Changzhou College of Information Technology, Changzhou 213164, China.
- School of Computer Science, Xi'an Jiaotong University, Xi'an 710049, China.
ORCID-Profile: Xiong Yang, Hao Wang, Jiong Ni
Förderung
- the Artificial Intelligence Empowered Curriculum and 471 Textbook System Construction Project for Higher Vocational Education of the Ministry of Education,
- Artificial Intelligence Empowered Curriculum and Textbook System Construction Project for Higher Vocational Education of the Ministry of Education
- Industry-University-Research Collaboration Project of the Science and Technology Department of Jiangsu Province
- Scientific Research Startup Fund of Changzhou College of Information Technology
Zitieren
APA
Yang, X., Wang, H., & Ni, J. (2026). VisualRNet: Lightweight Camera Rotation Estimation from Low-Resolution Optical Flow via Cross-Modal Supervision. Sensors, 26(9), 2655. https://doi.org/10.3390/s26092655
Vancouver
Yang X, Wang H, Ni J. VisualRNet: Lightweight Camera Rotation Estimation from Low-Resolution Optical Flow via Cross-Modal Supervision. Sensors. 2026;26(9):2655. doi:10.3390/s26092655
BibTeX
@article{Yang2026,
author = {Yang, Xiong and Wang, Hao and Ni, Jiong},
title = {VisualRNet: Lightweight Camera Rotation Estimation from Low-Resolution Optical Flow via Cross-Modal Supervision},
journal = {Sensors},
year = {2026},
volume = {26},
number = {9},
pages = {2655},
doi = {10.3390/s26092655}
}
EUVIMED