Home

Awesome

SonarSAM

This study presents the introduction of the Segment-Anything-Model (SAM) to sonar images. We conduct a comprehensive investigation into fine-tuning methods for SAM, including LoRA and visual prompt tuning. To facilitate comparison, we provide a framework that integrates these fine-tuning methods for SAM. If this project is helpful to your research, please consider citing our paper PDF.

@article{wang2023sonarsam,
  title={When SAM Meets Sonar Images},
  author={Wang, Lin and Ye, Xiufen and Zhu, Liqiang and Wu, Weijie and Zhang, Jianguo and Xing, Huiming and Hu, Chao},
  journal={arXiv preprint arXiv:2306.14109},
  year={2023}
}

Update

Dataset

The Marine Debris dataset is used in this work, which is available at Forward-Looking Sonar Marine Debris Datasets.

Training

python train_SAM_box.py --config ./configs/sam_box.yaml
python train_SAM.py --config ./configs/sam.yaml

License

The model is licensed under the Apache 2.0 license.

Acknowledgment

This project was developed based on the following awesome codes.