If you want to test our models remotely, run this colab notebook. Otherwise, you have to download our codebase and install it locally.

This framework is built using Python 3.9 and relies on the PyTorch 1.4.0+. The following command installs all necessary packages:

pip3 install -r requirements.txt

If you want to run training or testing, you must configure the paths to the datasets in config.yml.

$ ./run_demo.sh

First, download the datasets and pretrained weights and run the following code for evaluation:

python scripts/evaluate_model.py NoBRS \
--gpu 0 \
--checkpoint=./weights/imagenet21k_pretrain_cocolvis_finetune_segformerb5_epoch_54.pth \
--dataset=OAIZIB

Train the Swin-B model on the OAIZIB dataset.

python train.py models/iter_mask/swinformer_large_oaizib_itermask.py \
--batch-size=22 \
--gpu=0

We released two models: Swin-B and HRNet32 that can be downloaded in the release page.

The code is released under the MIT License. It is a short, permissive software license. Basically, you can do whatever you want as long as you include the original copyright and license notice in any copy of the software/source.

@article{liu2021isegformer,
  title={iSegFormer: Interactive Image Segmentation via Transformers with Application to 3D Knee MR Images},
  author={Liu, Qin and Xu, Zhenlin, and Jiao, Yining and Niethammer, Marc},
  journal={arXiv preprint arXiv:2112.11325},
  year={2021}
}