DiffDet4SAR: Diffusion-based Aircraft Target Detection Network for SAR Images

👑DiffDet4SAR is the first work of diffusion model for SAR image aircraft target detection.

DiffDet4SAR: Diffusion-based Aircraft Target Detection Network for SAR Images
accepted by GRSL DOI: 10.1109/LGRS.2024.3386020

🛠️ Updates

• (04/2024) Code is released.

🕸️ Dataset

SAR-AIRcraft1.0 (doi: 10.12000/JR23043)

📽️ Getting Started

The installation instruction and usage are in Getting Started with DiffusionDet.

🚉 Train/Evalution:

1. modifying the weight in DiffusionDet-main/configs/Base-DiffusionDet.yaml (use pre-train res50)
2. modifying the weight in DiffusionDet-main/configs/diffdet.coco.res50.300boxes.yaml
3. modifying DiffusionDet-main/detectron2/engine/defaults.py and the 98-122 line to your root.
4. As for other configs and their meaning, DifffusionDet is introduced in detail.
5. ATTENTION：In order to use the code directly and reduce the complexity of the code, I changed the images and annotations of the aircraft dataset into the coco format, and put them in the folder named coco.
6. 

🏝️ Quantative Results:

Quantitative results of different models evaluated by AP@50.

*The overall repository style is highly borrowed from DifffusionDet. Thanks to Shoufa Chen.

License

This project is under the CC-BY-NC 4.0 license. See LICENSE for details.

💡 Citing DiffDet4SAR

If you find DiffDet4SAR helpful to your research or wish to refer to the baseline results published here, please use the following BibTeX entry.

@ARTICLE{10494361,
  author={Zhou, Jie and Xiao, Chao and Peng, Bo and Liu, Zhen and Liu, Li and Liu, Yongxiang and Li, Xiang},
  journal={IEEE Geoscience and Remote Sensing Letters}, 
  title={DiffDet4SAR: Diffusion-Based Aircraft Target Detection Network for SAR Images}, 
  year={2024},
  volume={21},
  number={},
  pages={1-5},
  keywords={Aircraft;Object detection;Radar polarimetry;Feature extraction;Scattering;Noise;Convolution;Aircraft target detection;diffusion model;synthetic aperture radar (SAR)},
  doi={10.1109/LGRS.2024.3386020}}

Please light up the STAR⭐⭐⭐⭐⭐ to encourage more and more opensource on SAR image interpretations!

🥰🥳🥂