RL-Assisted SAM Segmentation in Diverse Medical Imaging Modalities

Code for the paper [Bridging the Gap in Lesion Segmentation: RL-Assisted SAM Segmentation in Diverse Medical Imaging Modalities]

Dependencies

• python 3.6.5
• numpy 1.14.5
• scipy 1.1.0
• Pytorch 0.4.0

Scripts

The folder 'scripts' contains the different bash scripts that could be used to train the same models used in the paper, for both Camvid and Cityscapes datasets.

• launch_baseline.sh: To train the baselines 'random'.
• launch_train_ralis.sh: To train the 'SAM RL agent' model.
• launch_test_ralis.sh: To test the 'SAM RL agent' model.