|
|
|
|
|
|
Introduction | GLH-Bridge Dataset | Benchmark | Generalization experiments | Citation | Contact
Bridges represent critical infrastructure components, serving as fundamental transportation facilities that traverse various landscapes. They hold substantial significance in the domains like civil transportation and disaster relief efforts. To ensure the visibility and integrity of bridges, it is essential to perform holistic bridge detection in large-size very-high-resolution (VHR) remote sensing images (RSIs). However, large-scale datasets and methods suitable for holistic bridge detection in large-sisze RSIs have yet to be explored. To ameliorate the scarcity of bridge datasets, we propose a large-scale dataset named GLH-Bridge comprising 6,000 VHR RSIs sampled from diverse geographic locations across the globe. These images encompass a wide range of sizes, varying from 2,048 x 2,048 to 16,384 x 16,384 pixels, and collectively feature 59,737 bridges. Moreover, we establish a large-scale benchmark including the OBB and HBB bridge detection tasks and propose the HBD-Net for holistic bridge detection. Finally, cross-dataset generalization experiments on two publicly available datasets illustrate the strong generalization capability of the GLH-Bridge dataset.
To fill the lack of pertinent bridge datasets, we propose a large-scale dataset named GLH-Bridge comprising 6,000 VHR RSIs sampled from diverse geographic locations across the globe, with the image sizes varying from 2,048 × 2,048 pixels to 16,384 × 16,384 pixels. Within GLH-Bridge, 59,737 bridges span diverse backgrounds like vegetation, dry riverbeds, and roads. Each of them has been manually annotated using both an oriented bounding box (OBB) and a horizontal bounding box (HBB).
GLH-Bridge dataset has six remarkable and important advantages:
Based on MMRotate and MMDetection, we conduct a comparative evaluation covering 18 advanced object detection methods in the GLH-Bridge dataset. Considering the characteristics of bridges, unlike the classic method of dividing labels based on area, we divide the labels based on length into short, middle, large, and huge, corresponding to pixel lengths of {(0,50], (50,200], (200,800], (800,16384]}. Our proposed HBD-Net achieves effective performance on the benchmark, particularly in the detection of huge bridges. The code of benchmark can be found in STAR MMRotate.
If you find this work helpful for your research, please consider citing our paper:
@article{li2024learning, title={Learning to Holistically Detect Bridges From Large-Size VHR Remote Sensing Imagery}, author={Li, Yansheng and Luo, Junwei and Zhang, Yongjun and Tan, Yihua and Yu, Jin-Gang and Bai, Song}, journal={IEEE Transactions on Pattern Analysis and Machine Intelligence}, volume={44}, number={11}, pages={7778--7796}, year={2024}, publisher={IEEE}} @article{li2024scene, title={STAR: A First-Ever Dataset and A Large-Scale Benchmark for Scene Graph Generation in Large-Size Satellite Imagery}, author={Li, Yansheng and Wang, Linlin and Wang, Tingzhu and Yang, Xue and Luo, Junwei and Wang, Qi and Deng, Youming and Wang, Wenbin and Sun, Xian and Li, Haifeng and Dang, Bo and Zhang, Yongjun and Yu, Yi and Yan Junchi}, journal={arXiv preprint arXiv:2406.09410}, year={2024}}
If you have any problems or feedback with the project, please contact: luojunwei@whu.edu.cn