SAR Ship Detector Using Cross-stage Feature Fusion and Decoupled Head with Mutual Guidance.
SAR Ship Detector Using Cross-stage Feature Fusion and Decoupled Head with Mutual Guidance, IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP 2025),--Accepted.
We have used the following versions of OS and softwares:
- OS: Windows10
- Python: 3.8
- GPU: RTX3060Ti
- CUDA: 11.2
- PyTorch: 1.9.0+cu111
- TorchVision: 0.10.0+cu111
- TorchAudio: 0.9.0
- MMCV-FULL: 1.3.17
- MMDetection: 2.20.0
conda create -n csff python=3.8
conda activate csffb. Install pytorch, torchvision and torchaudio following the official instructions, e.g.,
pip install torch==1.9.0+cu111 torchvision==0.10.0+cu111 torchaudio==0.9.0 -f https://download.pytorch.org/whl/torch_stable.htmlpip install openmim
mim install mmcv-full==1.3.17d. Install COCOAPI-AITOD for Evaluating on AI-TOD dataset(The VisDrone2019 dataset does not require it)
pip install "git+https://github.com/jwwangchn/cocoapi-aitod.git#subdirectory=aitodpycocotools"You can also refer to official instruction for installing COCOAPI-AITOD.
git clone https://github.com/SYLan2019/CSFF-MGDH.git
cd CSFF-MGDH
pip install -r requirements.txt
pip install -v -e .
# or "python setup.py install"Our data folder structure is as follows:
SOD-DEDDH
├── mmdet
├── tools
├── configs
├── data
│ ├── coco
│ │ ├── annotations
│ │ │ │─── train.json
│ │ │ │─── test.json
│ │ ├── images
| | | |___train
| | | | |─── ***.jpg
| | | | |─── ***.jpg
│ │ │ │─── test
| | | | |─── ***.jpg
│ │ │ │ |─── ***.jpgIf your data folder structure is different, you may need to change the corresponding paths in config files (configs/_base_/datasets/coco_detection.py).
The SOD-DEDDH's config files are in .
Please see MMDetection full tutorials with existing dataset for beginners.
python tools/train.py configs/CSFF_MGDH/python tools/test.py configs/CSFF_MGDH/ your_training_weight.pth --eval bbox