Skip to content

Latest commit

 

History

History
76 lines (55 loc) · 2.79 KB

README.md

File metadata and controls

76 lines (55 loc) · 2.79 KB

Weakly Supervised Building Extraction from High-Resolution Remote Sensing Images based on Building-Aware Clustering and Activation Refinement Network(TGRS)

Official pytorch implementation of "Weakly Supervised Building Extraction from High-Resolution Remote Sensing Images based on Building-Aware Clustering and Activation Refinement Network"

Updates

26 Aug, 2024: Initial upload

Requirement

  • This code is tested on Ubuntu 20.04, with Python 3.6, PyTorch 1.7.1, and CUDA 11.3.

Dataset & pretrained checkpoint

  • Prepare dataset, image-level labels, and pretrained checkpoints
    • Example directory hierarchy
    BAC-AR-Net
|--- Dataset
|    |--- RRDSD
|    |        |---JPEGImages
|    |        |---SegmentationClassAug
|--- weights
|    |--- RepVGG-B1g2-train.pth
|--- sess_rrdsd
|    |--- repvgg_cam_delpoy.pt
|--- result_rrdsd
|    |--- cam
|    |--- amn_cam
|    | ...
| ...

Execution

Pseudo-mask generation

  • Execute the bash file. 
    # Please see these files for the detail of execution.
bash script/generate_pseudo_mask.sh

Segmentation network

Fort the segmentation network, we experimented with DeepLab-V3+ based on PaddleSeg.

Acknowledgement

This code is highly borrowed from IRN, AMN, LPCAM. Thanks to Jiwoon, Ahn, Minhyun Lee, Zhaozheng Chen.

The codes for our previous work, including ACGC and MFR-PGC-Net, are also avaliable. Detailed information can be found in the respective publication papers.

Citation

If you find this work useful for your research, please cite our paper:

@ARTICLE{10623252,
  author={Zheng, Daoyuan and Wang, Shaohua and Feng, Haixia and Wang, Shunli and Ai, Mingyao and Zhao, Pengcheng and Li, Jiayuan and Hu, Qingwu},
  journal={IEEE Transactions on Geoscience and Remote Sensing}, 
  title={Weakly Supervised Building Extraction From High-Resolution Remote Sensing Images Based on Building-Aware Clustering and Activation Refinement Network}, 
  year={2024},
  volume={62},
  number={},
  pages={1-15},
  keywords={Buildings;Feature extraction;Semantics;Accuracy;Annotations;Semantic segmentation;Optimization;Building extraction;high-resolution (HR) remote sensing (RS) images;image-level labels;weakly supervised semantic segmentation},
  doi={10.1109/TGRS.2024.3438248}}