Install Requirement:
    pip install -r requirements.txt

Run Script:
    i) resize dataset images
        python downscale_image_script.py

    ii) generate training script
        python getnerate_run_script.py

    iii) run train_eval script
         inside ./run folder, run `bash run.sh`

    iv) plot result
        python plot_result.py

Training and Evaluation:
    see train_eval.ipynb

Resizing Dataset Images:
    see resize_dataset_image.py

Transfer .ipynb to .py:
    jupyter nbconvert --to python xxx.ipynb

Checking your jobs status on SCC:
    using following command:
        watch -n 5 "qstat | grep <your_username>"

TODO list:
- [x] Add Argparse to resizing_dataset_image.py and train_eval.ipynb
- [x] stratified when splitting dataset into train and val
- [x] add early stopping + patience variable
- [x] + plot_result.ipynb
- [] + logging
- [] + image_postfix arg in command line
- [] + N_EPOCH_FOR_GRADIENT_ACCUMULATION (EFFECTIVE_BATCH_SIZE = BATCH_SIZE * N_EPOCH_FOR_GRADIENT_ACCUMULATION)
- [] Try Efficient Net
- [] Try CS640 Network
- [] Try Unet + Transformer