Please refer to SLiMe(https://github.com/aliasgharkhani/SLiMe) for environment configuration details.

1. Download the corresponding dataset: iFSS you can follow the https://github.com/fcdl94/FSS and CD-FSS you can follow the https://github.com/slei109/PATNet?tab=readme-ov-file.

2. organize images and masks following this structure:

   pascal-5/
   	├──train/
   		├── 0/                                     
   		|   ├── images
   		|   └── masks
   		├── 1/                                   
   		├── ... 
   		├── 3/
   	├──val/
   coco-20/
   	├──train/
   		├── 0/                                     
   		|   ├── images
   		|   └── masks
   		├── 1/                                   
    	├── ... 
   		├── 3/
   	├──val/
   Deepglobe/                                        
   	├── 1/                                     
   	|   ├── images
   	|   └── masks
   	├── 2/                                   
   	├── ... 
   	├── 6/
   ISIC/                                        
   	├── 1/                                     
   	|   ├── images
   	|   └── masks
   	├── 2/                                   
   	├── ... 
   	├── 3/
   Lung/
   	├── train/                                     
   	|   ├── images
   	|   └── masks
   FSS-1000/
   	├── ab_wheel/
   	└── ...
   

3. Run the script:

   python process_data.py
   

Execute the training script:

python train.py

Run the testing script: {CHECKPOINT_DIR} is the path where your optimized embeddings are stored, {TEST_DIR} is the test data path, and {OUTPUT_DIR} is the path your want to output results.

python -m src.main2 --dataset pascal \ 
					--checkpoint_dir {CHECKPOINT_DIR} \
					--test_data_dir {TEST_DIR} \
					--output_dir {OUTPUT_DIR} \
					--save_test_predictions

The implementation is based on https://github.com/aliasgharkhani/SLiMe