UNet_for_SKR_LFE

Full code for the preparation of training and test set, as well as implementation of U-Net for semantic segmentation of Low Frequency Extensions of Saturn Kilometric Radiation.

Download the code and store according to:

• Folder
  
  • UNET_for_SKR_LFE
    • 0_prepare_data.py
    • 1_prepare_data.py
    • 2_prepare_data.py
    • 3_prepare_data.py
    • 4_prepare_data.py
    • 5_prepare_data.py
    • UNET.ipynb
  • input_data
    • SKR_2004_CJ.sav
    • ...
    • SKR_2017_001-258_CJ.sav
    • 2004_FGM_KRTP_1M.TAB
    • ...
    • 2017_FGM_KRTP_1M.TAB
    • 2004_FGM_KSM_1M.TAB
    • ...
    • 2004_FGM_KSM_1M.TAB
  • output_data
    • ML_lfes.json
  • ML_Dataset
    • flux_images
    • pol_images
    • mask_images

Description of files

Need to run prepare data files in order and change variable root to path to Folder at the top of each script.

0_prepare_data.py

Compile trajectory data from given files and store as single .csv file for each year in output_data folder.

1_prepare_data.py

Separate data with LFEs fully labelled and extract empty intervals (without LFE) into 5 hour chunks. Save .csv file of start and end times of both LFEs and non-LFEs along with class label to output_data.

2_prepare_data.py

Plot and save spectrogram images and corresponding binary mask for each labelled instance of LFE/Non-LFE.

3_prepare_data.py

Perform data augmentation and plot and save spectrogram images and corresponding binary mask.

4_prepare_data.py

Save .csv file with start and end times of LFEs, Non-LFEs and augmented data along with corresponding latitude median and standard deviation and local time median and standard deviation over each interval.

5_prepare_data.py

Separate data into train and test, and save images, masks and labels to folders corresponding to index of each image.

UNET.ipynb

Implementation of modified U-NET.