S-SAM

This repository contains the code for Low-Rank Adaptation of Segment Anything Model for Surgical Scene Segmentation

Environment File

Create a new conda environment with the config file given in the repository as follows:

conda env create -f ssam_env.yml
conda activate s-sam

General file descriptions

• data_transforms/*.py - data transforms defined here for different datasets.
• data_utils.py - functions to generate dataloaders for different datasets
• model.py - model architectures defined here
• prompt_adapted_segment_anything/modeling/svd_layers.py - code for the singular value tuning modifications used in the model
• train.py - code for general training, common to all datasets
• driver_scratchpad.py - driver code for training models.
• eval/*/generate_predictions.py - code for generating results for a given dataset
• eval/*/generate_predictions.sh - script to run generate_predictions for generating results for all labels of interest.
• model_svdtuning.yml - config file for defining various model hyperparameters for SVDSAM
• config_<dataset_name>.yml - config file for defining various dataset related hyperparameters

Example Usage for Training

python driver_scratchpad.py --model_config model_svdtuning.yml --data_config config_cholec8k.yml --save_path "./temp.pth"

Please refer to driver_scratchpad.py for other command line options and parameters.

Example Usage for Evaluation

cd eval/cholec8k

bash generate_predictions_cholec.sh

Citation

To be added

Please feel free to reach out to me or raise an issue in case of trouble while running the code.