UMFNet: Frequency-Guided Multi-Scale Fusion with Dynamic Noise Suppression for Robust Low-Light Object Detection
Source abstract
Abstract
The dominant low-light object detectors face the following spectral trilemma: (1) the loss of high-frequency structural details during denoising, (2) the amplification of low-frequency illumination distortion, and (3) cross-band interference in multi-scale features. To resolve these intertwined challenges, we present UMFNet—a frequency-guided detection framework that unifies adaptive frequency distillation with inter-band attention coordination. Our technical breakthroughs manifest through three key innovations: (1) a frequency-adaptive fusion (FAF) module employing learnable wavelet kernels (16–64 decomposition basis) with dynamic SNR-gated thresholding, achieving an 89.7% photon utilization rate in ≤1 lux conditions—2.4× higher than fixed-basis approaches; (2) a spatial-channel coordinated attention (SCCA) mechanism with dual-domain nonlinear gating that reduces high-frequency hallucination by 37% through parametric phase alignment (verified via gradient direction alignment coefficient ρG = 0.93); (3) a spectral perception loss combining the frequency-weighted structural similarity index measure (SSIM) with gradient-aware focal modulation, enforcing physics-constrained feature recovery. Extensive validation demonstrates UMFNet’s leadership: 73.1% mAP@50 on EXDark (+6.4% over YOLOv8 baseline), 58.7% on DarkFace (+3.1% over GLARE), and 40.2% on thermal FLIR ADAS (+9.7% improvement). This work pioneers a new paradigm for precision-critical vision systems in photon-starved environments.
Source: crossref · Quality: baseline
Research context
OpenAlex topics: Infrared Target Detection Methodologies · Image Enhancement Techniques · Industrial Vision Systems and Defect Detection
SciVX discovery tags are terminology-based. OpenAlex topics are source-provided classifications.
Affiliations
School of Computer Science, China University of Geosciences, Wuhan 430074, China
Informatization Office, China University of Geosciences, Wuhan 430074, China
References (31)
Show source-provided references
1. Getting to Know Low-light Images with The Exclusively Dark Dataset10.1016/j.cviu.2018.10.010 ↗
2. Cui, Z., Li, K., Gu, L., Su, S., Gao, P., Jiang, Z., Qiao, Y., and Harada, T. (2022). You only need 90k parameters to adapt light: A light weight transformer for image enhancement and exposure correction. arXiv.
3. Guo, C.G., Li, C., Guo, J., Loy, C.C., Hou, J., Kwong, S., and Cong, R. (2020, January 14–19). Zero-reference deep curve estimation for low-light image enhancement. Proceedings of the IEEE conference on computer vision and pattern recognition (CVPR), Seattle, WA, USA.10.1109/CVPR42600.2020.00185 ↗
4. Advancing Image Understanding in Poor Visibility Environments: A Collective Benchmark Study10.1109/TIP.2020.2981922 ↗
5. Wang, Y., Wan, R., Yang, W., Li, H., Chau, L.P., and Kot, A.C. (2021). Low-Light Image Enhancement with Normalizing Flow. arXiv.10.1609/aaai.v36i3.20162 ↗
6. Girshick, R. (2015, January 7–13). Fast R-CNN. Proceedings of the 2015 IEEE International Conference on Computer Vision (ICCV), Santiago, Chile.10.1109/ICCV.2015.169 ↗
7. Abdulla, W. (2024, October 15). Mask R-CNN for Object Detection and Instance Segmentation on Keras and TensorFlow. Available online: https://github.com/matterport/Mask_RCNN.
8. Girshick, R., Donahue, J., Darrell, T., and Malik, J. (2014). Rich feature hierarchies for accurate object detection and semantic segmentation. arXiv.10.1109/CVPR.2014.81 ↗
9. Ren, S., He, K., Girshick, R., and Sun, J. (2016). Faster R-CNN: Towards Real-Time Object Detection with Region Proposal Networks. arXiv.10.1109/TPAMI.2016.2577031 ↗
10. Zhang, Y., Zhang, J., and Guo, X. (2019, January 21–25). Kindling the Darkness: A Practical Low-light Image Enhancer. Proceedings of the 27th ACM International Conference on Multimedia, MM’19, New York, NY, USA.10.1145/3343031.3350926 ↗
11. Enlightengan: Deep light enhancement without paired supervision10.1109/TIP.2021.3051462 ↗
12. Liu, W., Ren, G., Yu, R., Guo, S., Zhu, J., and Zhang, L. (March, January 22). Image-Adaptive YOLO for Object Detection in Adverse Weather Conditions. Proceedings of the AAAI Conference on Artificial Intelligence, Online.
13. Cui, Z., Qi, G.J., Gu, L., You, S., Zhang, Z., and Harada, T. (2022). Multitask AET with Orthogonal Tangent Regularity for Dark Object Detection. arXiv.10.1109/ICCV48922.2021.00255 ↗
14. Chen, Y., Wang, B., Guo, X., Zhu, S., He, J., Liu, X., and Yuan, J. (2024, January 1–5). DEYOLO: Dual-Feature-Enhancement YOLO for Cross-Modality Object Detection. Proceedings of the International Conference on Pattern Recognition, Kolkata, India.10.1007/978-3-031-78447-7_16 ↗
15. Redmon, J. (2018). Yolov3: An incremental improvement. arXiv.
16. Redmon, J., Divvala, S., Girshick, R., and Farhadi, A. (2016). You Only Look Once: Unified, Real-Time Object Detection. arXiv.10.1109/CVPR.2016.91 ↗
17. Wang, J., Yang, P., Liu, Y., Shang, D., Hui, X., Song, J., and Chen, X. (2023). Research on Improved YOLOv5 for Low-Light Environment Object Detection. Electronics, 12.10.3390/electronics12143089 ↗
18. Yin, X., Yu, Z., Fei, Z., Lv, W., and Gao, X. (2023). PE-YOLO: Pyramid Enhancement Network for Dark Object Detection. arXiv.10.1007/978-3-031-44195-0_14 ↗
19. Wei, C., Fan, H., Xie, S., Wu, C.Y., Yuille, A., and Feichtenhofer, C. (2023). Masked Feature Prediction for Self-Supervised Visual Pre-Training. arXiv.10.1109/CVPR52688.2022.01426 ↗
20. Fujieda, S., Takayama, K., and Hachisuka, T. (2018). Wavelet Convolutional Neural Networks. arXiv.
21. Xuan, H., Yang, B., and Li, X. (2025). Exploring the Impact of Temperature Scaling in Softmax for Classification and Adversarial Robustness. arXiv.
22. A Computational Approach to Edge Detection10.1109/TPAMI.1986.4767851 ↗
23. Li, Y., Yuan, G., Wen, Y., Hu, J., Evangelidis, G., Tulyakov, S., Wang, Y., and Ren, J. (2022). EfficientFormer: Vision Transformers at MobileNet Speed. arXiv.
24. Lv, F., Lu, F., Wu, J., and Lim, C. (2018, January 3–6). MBLLEN: Low-light Image/Video Enhancement Using CNNs. Proceedings of the British Machine Vision Conference (BMVC), Newcastle, UK.
25. Zhao, Y., Lv, W., Xu, S., Wei, J., Wang, G., Dang, Q., Liu, Y., and Chen, J. (2024). DETRs Beat YOLOs on Real-time Object Detection. arXiv.10.1109/CVPR52733.2024.01605 ↗
26. Wang, C.Y., Yeh, I.H., and Liao, H.Y.M. (2024). YOLOv9: Learning What You Want to Learn Using Programmable Gradient Information. arXiv.10.1007/978-3-031-72751-1_1 ↗
27. Hnewa, M., and Radha, H. (2021, January 9–22). Multiscale Domain Adaptive Yolo For Cross-Domain Object Detection. Proceedings of the 2021 IEEE International Conference on Image Processing (ICIP), Anchorage, AK, USA.10.1109/ICIP42928.2021.9506039 ↗
28. Zhang, S., Tuo, H., Hu, J., and Jing, Z. (2021). Domain Adaptive YOLO for One-Stage Cross-Domain Detection. arXiv.
29. Zhang, X., Zeng, H., Guo, S., and Zhang, L. (2022). Efficient Long-Range Attention Network for Image Super-resolution. arXiv.10.1007/978-3-031-19790-1_39 ↗
30. Wei, C., Wang, W., Yang, W., and Liu, J. (2018). Deep retinex decomposition for low-light enhancement. arXiv.
31. Yu, F., Chen, H., Wang, X., Xian, W., Chen, Y., Liu, F., Madhavan, V., and Darrell, T. (2020, January 14–19). BDD100K: A Diverse Driving Dataset for Heterogeneous Multitask Learning. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), Seattle, WA, USA.10.1109/CVPR42600.2020.00271 ↗