Update README.md

code/processing/README.md

@@ -33,6 +33,16 @@ model.
       wget https://isic-challenge-data.s3.amazonaws.com/2018/ISIC2018_Task1_Training_GroundTruth.zip
       unzip ISIC2018_Task1_Training_GroundTruth.zip
       ```
+- You need .txt files that lists the paths for the train, validation, and test splits of the image dataset. You can either download these text files from hugging face or generate them yourself. These files must be placed under `../../data/dataset_splits/`.
+
+   - In order to use the pre-generated dataset split text files (e.g. ```all_tones_real_ISIC_1.0_add_synth_0.2```): 
+      - Download ```all_tones_real_ISIC_1.0_add_synth_0.2``` from hugging face and place within Github repo:
+          ```
+          cd $CWD/code/processing
+          NAME="all_tones_real_ISIC_1.0_add_synth_0.2"
+          python download_split.py --name $NAME --saveDir '../../'
+          ``` 
+   - To generate the dataset split text files, follow the instructions provided under “Split the Datasets into Training, Validation, and Test Sets.”
 
 ## Training/Testing Segmentation Model
 This code below is heavily based on the [DermsegDiff](https://github.com/xmindflow/DermoSegDiff) implementation of [1].
@@ -73,12 +83,7 @@ jupyter notebook visualization.ipynb
 ```
 
 ### Training Segmentation Model (DermoSegDiff):
-- Download ```all_tones_real_ISIC_1.0_add_synth_0.2``` from hugging face and place within Github repo:
-    ```
-    cd $CWD/code/processing
-    NAME="all_tones_real_ISIC_1.0_add_synth_0.2"
-    python download_split.py --name $NAME --saveDir '../../'
-    ```
+- Make sure the dataset split text files are located in `../../data/dataset_splits/`, as outlined in the "Setup" section.
 
 - Run training: 
     ```
@@ -104,14 +109,14 @@ jupyter notebook visualization.ipynb
      
       ```
       cd $CWD/code/processing
-      python -u data_split_ISIC.py --real_data_ratio 1.0 --synth_data_ratio 0.2 --saveDir '../../data/outputs/dataset_splits/' --add_synth 
+      python -u data_split_ISIC.py --real_data_ratio 1.0 --synth_data_ratio 0.2 --saveDir '../../data/dataset_splits/' --add_synth 
       ```
 
    - Run `data_split_HAM.py` to generate own synthetic/HAM10k data splits:
      
       ```
       cd $CWD/code/processing
-      python -u data_split_HAM.py --real_data_ratio 1.0 --synth_data_ratio 0.2 --saveDir '../../data/outputs/dataset_splits/' --add_synth
+      python -u data_split_HAM.py --real_data_ratio 1.0 --synth_data_ratio 0.2 --saveDir '../../data/dataset_splits/' --add_synth
       ```
       
       These scripts accept the following arguments: