EfficientViT-Based Adaptive Feature Normalization for Cross-Domain UAV Object Detection

Reliable small-object detection in unmanned aerial vehicle (UAV) imagery is difficult on resource-constrained platforms, and lightweight detectors often degrade when applied to data captured under different sensors, altitudes, and environmental conditions. This study investigates a lightweight framework intended to improve cross-domain robustness, combining an EfficientViT backbone with Adaptive Feature Normalization (AFN), which learns a per-channel blend of Batch and Instance Normalization. The framework also uses MixStyle for domain generalization, BiFPN-inspired multi-scale fusion, and a task-decoupled detection head. The model contains 4.48 million parameters and requires 22.73 GFLOPs at 640×640 input. Training followed a source-only, zero-shot protocol, with VisDrone2019 as the source domain and UAVDT as the target, under a two-class unified vehicle taxonomy. Evaluation covers detection accuracy (mAP), parameter count, and zero-shot cross-domain transfer, with YOLOv8n and RTMDet-tiny as baselines. After a non-finite-loss training defect was traced to DataLoader worker configuration and resolved, single-domain training reached an mAP@0.5 of 0.0385 on VisDrone2019. Ablation results identified the task-decoupled head as the most influential component. In zero-shot VisDrone2019→UAVDT transfer, MixStyle did not improve on the baseline: the two-seed mean mAP@0.5 was 0.0133 for the baseline and 0.0061 with MixStyle, although its effect was small relative to seed-to-seed variance. Under a matched protocol, the proposed model exceeded RTMDet-tiny (mAP@0.5 of 0.0238) but fell well short of YOLOv8n (0.2259). Limitations include the absence of data augmentation, unmeasured inference speed for the proposed model, and an unexecuted DroneVehicle evaluation. The study provides a transparent empirical baseline and identifies priorities for future cross-domain UAV detection research.

UAV small object detection, EfficientViT, adaptive feature normalization, cross-domain generalization, MixStyle

1. Cai, L., & Li, S. (2026). DSPE-ViT: a lightweight vision transformer with dynamic sparse positional encoding for dense small object detection in UAV imagery. Frontiers in Neurorobotics, 20, 1849093. [Google Scholar] [Crossref]

2. Chen, K. C., Malligere Shivanna, V., & Guo, J. I. (2026). YOLO-DAA: Directional Area Attention for Lightweight Tiny Object Detection in Maritime UAV Imagery. Drones, 10(4), 283. [Google Scholar] [Crossref]

3. Di, J., Wang, Q., & Yang, Y. (2026). CMM-DETR: an efficient transformer-based detector for remote sensing measurement applications. Measurement Science and Technology, 37(19), 195401. [Google Scholar] [Crossref]

4. Dong, C., Wang, Y., Zhai, Y., Li, Y., Liu, X., Mai, C., ... & Chen, C. P. (2026). DCA-YOLO: A Dynamic Cross-Attention Network for Small Vehicle Detection in UAV Aerial Imagery. IEEE Transactions on Aerospace and Electronic Systems. [Google Scholar] [Crossref]

5. Fan, X., Gao, M., Wang, Y., Tang, H., Sun, F., & Song, C. (2026). Data-Efficient Insulator Defect Detection in Power Transmission Systems via Multi-Granularity Feature Learning and Latent Context-Aware Fusion. Electronics, 15(5), 1081. [Google Scholar] [Crossref]

6. Li, C., Cao, J., Hao, Q., Song, C., Yao, H., & Wei, Z. (2026). RADFRNet: Detail-Enhanced Feature Recalibration for Infrared Small-Target Detection Based on an Improved YOLOv11. Sensors, 26(15), 4850. [Google Scholar] [Crossref]

7. Liu, J., Xie, Y., & Wang, J. (2026). A lightweight and robust framework for small object detection in UAV imagery. Computer Vision and Image Understanding, 267, 104735. [Google Scholar] [Crossref]

8. Liu, M., Hou, Y., Duan, X., Feng, B., Hou, H., Yang, M., ... & Gu, L. (2026). Multi-scale feature reconstruction detection transformer (MSFR-DETR): a real-time detector for small objects. Journal of Real-Time Image Processing, 23(3), 97. [Google Scholar] [Crossref]

9. Liu, S., Zhang, T., Ren, Y., Liu, Q., & Wang, L. (2026). A novel lightweight FusionMamba-YOLO algorithm with the hybrid Mamba-CNN architecture for real-time scene understanding in assistive robotic navigation. Complex & Intelligent Systems. [Google Scholar] [Crossref]

10. Liu, Z., Han, J., Zhai, G., & Li, W. (2026). LUW-DETR: a lightweight underwater object detection based on DETR. Journal of Real-Time Image Processing, 23(1), 14. [Google Scholar] [Crossref]

11. Lu, W., Niu, C., Liu, W., Lan, C., & Wang, S. (2025). Foundation Model-Based Auxiliary Framework for Object Detection in Aerial Remote Sensing Images. IEEE Transactions on Aerospace and Electronic Systems. [Google Scholar] [Crossref]

12. Mao, Y., Zhang, H., Li, R., Zhu, F., Sun, R., & Ji, P. (2025). HSF-DETR: Hyper scale fusion detection transformer for multi-perspective UAV object detection. Remote Sensing, 17(12), 1997. [Google Scholar] [Crossref]

13. Moskalenko, V. V., Moskalenko, A. S., Moskalenko, Y. V., & Vatsenko, A. V. (2026). Resource-efficient, robust, and adaptive object detection in UAV imagery. Radio Electronics, Computer Science, Control, (2), 85-99. [Google Scholar] [Crossref]

14. Ren, Y. (2026, April). Optimization Method for Small Object Detection in ViT-DETR Based on Deformable Cross-Attention and Multi-Scale Feature Alignment. In 2026 11th International Conference on Intelligent Computing and Signal Processing (ICSP) (pp. 1815-1818). IEEE. [Google Scholar] [Crossref]

15. Shi, C., Wang, H., Fu, X., Liu, P., & Ke, H. (2026). A Small Object Detection Transformer for UAV Remote Sensing Imagery via Multi-Scale Perception and Cross-Spatial-Frequency Domain Fusion. Remote Sensing, 18(9), 1394. [Google Scholar] [Crossref]

16. Tian, B., Hu, M., Li, L., Zhao, X., Tang, M., & Wang, Z. (2026). AMFNet: adaptive multi-scale fusion network for small object detection. Complex & Intelligent Systems. [Google Scholar] [Crossref]

17. Tu, X., Gao, Z., Wei, L., Zhang, Y., Liu, J., Yang, M., ... & Li, X. (2026). Small object detection in UAV aerial imagery through domain consistency optimization for cross-resolution semantic alignment. Pattern Recognition, 113463. [Google Scholar] [Crossref]

18. Wang, S., Guo, Z., Yu, Z., & Zhou, Z. (2026). CFP-DETR: Collaborative Feature Purification Network with Spatial Alignment for Aerial Small Object Detection. Remote Sensing, 18(11), 1750. [Google Scholar] [Crossref]

19. Xu, Z., Zhou, Y., Wen, S., & Jing, W. (2025). FIDC-YOLO: improved YOLO for detecting pine wilt disease in UAV remote sensing images via feature interaction and dependency capturing. IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, 18, 14674-14687. [Google Scholar] [Crossref]

20. Zhang, J., Zhong, H., Du, G., Zhang, B., & Zhang, H. (2026). Aerial small object detection via dynamic convolution and hierarchical attention fusion for UAV imagery. Scientific Reports. [Google Scholar] [Crossref]

21. Zhang, X., & Hu, X. (2026). Multi-stage feature purification for robust object detection in haze-degraded aerial images. Scientific Reports. [Google Scholar] [Crossref]

22. Zhou, S., Liu, M., Zhou, J., & Zheng, R. (2026). Vision transformer with salience self-attention for underwater and aerial object recognition and tracking. Neurocomputing, 132631. [Google Scholar] [Crossref]

23. Zuo, G., Wang, Y., Sha, Y., Xia, Y., Tang, M., Jia, H., & Chi, R. (2026). DCM-YOLO: Robust Electric Bicycle Detection in Confined Indoor Environments Under Occlusion and Image Degradation. Symmetry, 18(6), 1040. [Google Scholar] [Crossref]