EUVIMEDEUVIMED entdecken
Crossref

SFNet: Sparse Fusion Network for Object Detection of Cross-Modal Images

Haidong Xiao, Zhigang Ren, Ziyu Li, Zhuoxun Zeng, Shuangping Yang, Zimu Teng, Yijie Wang, Shengze Cai, Chao Xu, Zongze Wu

IEEE Sensors Journal · 15.06.2026 · 26 (12) · 18708-18719

DOI: 10.1109/jsen.2026.3689924 ↗

Kurz gefasst

Fachartikel in der Zeitschrift IEEE Sensors Journal, erschienen am 15.06.2026. Verfasst von Haidong Xiao und 9 weiteren Autor:innen. Die Arbeit nennt 55 Literaturangaben. Gefördert von: Basic and Applied Basic Research Foundation of Guangdong Province, Guangdong Major Project of Basic and Applied Basic Research, Key Research and Development Project of China National Tobacco Corporation.

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Effective aggregation of complementary information from visible (VIS) and infrared (IR) modalities can significantly enhance the performance and robustness of multimodal object detection systems. However, existing methods often suffer from limitations, such as overly complex architectures or ineffective cross-modal interaction at the semantic level. Furthermore, some approaches integrate features extensively without adequate filtering mechanisms, potentially introducing interference from redundant information. To address these challenges, we propose SFNet, an end-to-end cross-modal object detection network based on a sparse fusion strategy. SFNet utilizes a weight-shared twinned backbone to synchronously encode feature maps from both modalities. We introduce a novel sparse fusion module (SFM) that operates at the semantic level to refine salient cross-modal features while simultaneously filtering redundant components during information interaction. Additionally, an adaptive proportional modulation module (APMM) is incorporated to dynamically adjust attention weights based on the characteristics of the fused multimodal feature distribution. Extensive qualitative and quantitative experiments conducted on several benchmark datasets (M3FD, KAIST, FLIR) validate the effectiveness and superiority of SFNet. Our method achieves mean average precision (mAP) scores of 90.4% on the M3FD dataset, 78.3% on the KAIST dataset, and 87. 0% on the FLIR dataset.

Förderung

  • Basic and Applied Basic Research Foundation of Guangdong Province
  • Guangdong Major Project of Basic and Applied Basic Research
  • Key Research and Development Project of China National Tobacco Corporation
  • Open Research Fund from Guangdong Laboratory of Artificial Intelligence and Digital Economy [Shenzhen (SZ)].

Zitieren

APA

Xiao, H., Ren, Z., Li, Z., Zeng, Z., Yang, S., Teng, Z., Wang, Y., Cai, S., Xu, C., & Wu, Z. (2026). SFNet: Sparse Fusion Network for Object Detection of Cross-Modal Images. IEEE Sensors Journal, 26(12), 18708-18719. https://doi.org/10.1109/jsen.2026.3689924

Vancouver

Xiao H, Ren Z, Li Z, Zeng Z, Yang S, Teng Z, et al. SFNet: Sparse Fusion Network for Object Detection of Cross-Modal Images. IEEE Sensors Journal. 2026;26(12):18708-18719. doi:10.1109/jsen.2026.3689924

BibTeX
@article{Xiao2026,
  author = {Xiao, Haidong and Ren, Zhigang and Li, Ziyu and Zeng, Zhuoxun and Yang, Shuangping and Teng, Zimu and Wang, Yijie and Cai, Shengze and Xu, Chao and Wu, Zongze},
  title = {SFNet: Sparse Fusion Network for Object Detection of Cross-Modal Images},
  journal = {IEEE Sensors Journal},
  year = {2026},
  volume = {26},
  number = {12},
  pages = {18708--18719},
  doi = {10.1109/jsen.2026.3689924}
}
Kostenlos starten →