2023.07.14: The SPPNet model and training code have been submitted. The paper will be updated later.

2023.08.24: The paper has been accepted by MICCAI-MLMI 2023. The preprint has been available at arXiv.

2023.09.27: Release a New Beta version for users who want to fine-tune the SAM pre-trained image encoder. We add the adapter based on Medical-SAM-Adapter.

1. pytorch==1.10.0
2. pytorch-lightning==1.1.0
3. albumentations==0.3.2
4. seaborn
5. sklearn

NVIDIA RTX2080Ti Tensor Core GPU, 4-core CPU, and 28GB RAM

To apply the model on a custom dataset, the data tree should be constructed as:

    ├── data
          ├── images
                ├── image_1.png
                ├── image_2.png
                ├── image_n.png
          ├── masks
                ├── image_1.npy
                ├── image_2.npy
                ├── image_n.npy

python train.py --dataset your/data/path --jsonfile your/json/path --loss dice --batch 16 --lr 0.001 --epoch 50 

python eval.py --dataset your/data/path --jsonfile your/json/path --model save_models/model_best.pth --debug True

The codes are modified from SAM and MobileSAM.