Frequency-Aware Linear Attention for Small Object Detection in UAV Remote Sensing of River and Lake Shoreline Areas.
| File | Description |
|---|---|
CSP_FALA.py |
Frequency-Aware Linear Attention (FALA) module, replaces C3K2 in YOLO11 backbone and neck |
HAT.py |
Hybrid Attention Transformer (HAT) module for small target super-resolution enhancement |
hf-yolo-obb.yaml |
YOLO model configuration file defining the HF-YOLO network architecture |
FALAexp.py |
Visualization experiments: attention heatmap and frequency spectrum analysis |
Copy CSP_FALA.py and HAT.py to `ultralytics/nn/Extramodules/
The RSI (River Shoreline Inspection) dataset is a UAV-based aerial imagery dataset for targets along river and lake shorelines. It was collected to support automated inspection tasks in water resource management. The RSI Dataset used in this work can be downloaded from the Releases page. The dataset is split into 3 parts (rsi_dataset.zip.001-003); download all parts and extract to obtain the complete dataset. More details of this dataset can be found in our paper.
If you find this work useful, please cite our paper:
Bao, J., Zhang, Z., Petropoulos, G. P., Fang, Z., Shen, J., & Jing, L. (2026). HF-YOLO: An improved object detection algorithm for typical targets in river and lake shoreline areas. Remote Sensing Applications: Society and Environment, 43, 102168. https://doi.org/10.1016/j.rsase.2026.102168
@article{BAO2026102168,
title = {HF-YOLO: An improved object detection algorithm for typical targets in river and lake shoreline areas},
journal = {Remote Sensing Applications: Society and Environment},
volume = {43},
pages = {102168},
year = {2026},
issn = {2352-9385},
doi = {10.1016/j.rsase.2026.102168},
url = {https://www.sciencedirect.com/science/article/pii/S2352938526003010},
author = {Juncheng Bao and Zhen Zhang and George P. Petropoulos and Zihan Fang and Jiawang Shen and Li Jing},
keywords = {Small object detection, Linear attention mechanism, UAV remote sensing, Image frequency domain analysis, Shoreline monitoring}
}MIT