👑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
- (04/2024) Code is released.
SAR-AIRcraft1.0 (doi: 10.12000/JR23043)
The installation instruction and usage are in Getting Started with DiffusionDet.
- modifying the weight in DiffusionDet-main/configs/Base-DiffusionDet.yaml (use pre-train res50)
- modifying the weight in DiffusionDet-main/configs/diffdet.coco.res50.300boxes.yaml
- modifying DiffusionDet-main/detectron2/engine/defaults.py and the 98-122 line to your root.
- As for other configs and their meaning, DifffusionDet is introduced in detail.
- 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.
Quantitative results of different models evaluated by AP@50.
*The overall repository style is highly borrowed from DifffusionDet. Thanks to Shoufa Chen.
This project is under the CC-BY-NC 4.0 license. See LICENSE for details.
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!
🥰🥳🥂