Skip to main page content
U.S. flag

An official website of the United States government

Dot gov

The .gov means it’s official.
Federal government websites often end in .gov or .mil. Before sharing sensitive information, make sure you’re on a federal government site.

Https

The site is secure.
The https:// ensures that you are connecting to the official website and that any information you provide is encrypted and transmitted securely.

Access keys NCBI Homepage MyNCBI Homepage Main Content Main Navigation
. 2025 Oct 8;15(1):35083.
doi: 10.1038/s41598-025-18881-3.

CAFE-YOLO: an object detection algorithm from UAV perspective fusing channel attention and fine-grained feature enhancement

Chenglong Mi  1   2 , Yanling Chen  1   2 , Lei Zhu  2 , Xiaolong Yu  2 , Jinghua Sun  3 , Qianchun Luo  2
Affiliations

CAFE-YOLO: an object detection algorithm from UAV perspective fusing channel attention and fine-grained feature enhancement

Chenglong Mi et al. Sci Rep. .

Abstract

In aerial imagery captured by drones, object detection tasks often face challenges such as a high proportion of small objects, complex background interference, and insufficient lighting conditions, all of which substantially affect feature representation and detection accuracy. To address these challenges, a novel object detection algorithm named channel attention and fine-grained enhancement YOLO (CAFE-YOLO) is proposed. This algorithm incorporates a channel attention mechanism into the backbone network to enhance the focus on critical features while suppressing redundant information. Furthermore, a fine-grained feature enhancement module is introduced to extract local detail features, improving the perception of small and occluded objects. In the detection head, a lightweight attention-guided feature fusion strategy is designed to further optimize object localization and classification performance. Experimental results on the VisDrone2019 dataset show that the proposed method achieves significantly better detection performance than most existing advanced algorithms in complex drone-captured imaging scenarios. While maintaining a lightweight architecture, it reaches a mean average precision at IoU threshold 0.5 of 44.6%, demonstrating substantial improvements in both overall detection accuracy and robustness.

Keywords: Channel attention mechanism; Drone-captured imagery; Fine-grained feature enhancement; Object detection.

PubMed Disclaimer

Conflict of interest statement

Declarations. Competing interests: The authors declare no competing interests.

Figures

Figure 1
Figure 1
CAFE-YOLO model.
Figure 2
Figure 2
The structure of CAG.
Figure 3
Figure 3
The structure of FGFE.
Figure 4
Figure 4
The structure of EFF.
Figure 5
Figure 5
VisDrone dataset and DroneVehicle dataset.
Figure 6
Figure 6
Training process curve variation.
Figure 7
Figure 7
Visual comparison of detection results at different stages of the ablation study.
Figure 8
Figure 8
Comparison of performance across various models on the test set.
Figure 9
Figure 9
Under low-light conditions.
Figure 10
Figure 10
Under complex scene conditions.
Figure 11
Figure 11
Under occlusion conditions.
Figure 12
Figure 12
Under densely populated object scenarios.
See this image and copyright information in PMC

References

    1. Zhang, Y., Wu, C., Zhang, T., Liu, Y. & Zheng, Y. Self-Attention Guidance and Multiscale Feature Fusion-Based UAV Image Object Detection. IEEE Geosci. Remote Sens. Lett.20, 1–5 (2023).
    1. Wang, X., He, N., Hong, C., Wang, Q. & Chen, M. Improved YOLOXX Based UAV Aerial Photography Object Detection Algorithm. Image Vis. Comput.135, 104697 (2023).
    1. Zeng, S., Yang, W., Jiao, Y., Geng, L. & Chen, X. SCA-YOLO: A New Small Object Detection Model for UAV Images. Vis. Comput.40, 1787–1803 (2023).
    1. Gao, P. et al. “Double FCOS: A Two-Stage Model Utilizing FCOS for Vehicle Detection in Various Remote Sensing Scenes’’. IEEE J. Sel. Topics Appl. Earth Obs. Remote Sens15, 4730–4743 (2022).
    1. Huyan, N., Zhang, X., Quan, D., Chanussot, J. & Jiao, L. Cluster-Memory Augmented Deep Autoencoder via Optimal Transportation for Hyperspectral Anomaly Detection. IEEE Trans. Geosci. Remote Sens.60, 5531916 (2022).

LinkOut - more resources