Chorus wave segmentation in magnetic spectra

This is the repository containing the code for the paper titled ....

Citation TBD

1. To setup the environment:

    conda create --name chorus_wave --file conda_packages.txt &&
    conda activate chorus_wave
   

   Within conda env install the following pip packages:

   pip install pip_requirements.txt
   

2. Download weights for SAM model

   wget -P sam https://dl.fbaipublicfiles.com/segment_anything/sam_vit_h_4b8939.pth
   

3. Download the data and unzip the files in the data folder (For access to the full dataset. For the demo skip steps 3 and 4, and jump straight to step 5.)

   wget -P data/ http://babeta.ufa.cas.cz/dpisa/down/europlanet/chorus_part1.zip &&
   wget -P data/ http://babeta.ufa.cas.cz/dpisa/down/europlanet/chorus_part2.zip
   

4. Run data preparation:

   python dataprep.py
   

5. To download the data for demo: git lfs pull

6. Run the experiments in Jupyter notebooks under notebooks

Project Structure

├── (data)                                            # Data directory
│   ├── (npy_1)                                       # To include the unzipped data partition 1
│   ├── (npy_2)                                       # To include the unzipped data partition 2
│   ├── (processed)                                   # Directory to contain processed data images and corresponding annotations
│   │   └── (images)                                  # Includes images created from spectrograms using dataprep.py
│   │   └── (masks+)                                  # Includes the positive annotations created using 01-Annotation.ipynb
│   │   │   └── (train)                               # Train partition
│   │   │   └── (test)                                # Test partition
├── (notebooks)                                       # jupyter notebooks for various experiments
│   ├── 01-Annotation.ipyn                            # Annotate the images using SAM box prompt
│   ├── 02-Modelling-baseline-random_sampling.ipynb   # Baseline experiments with data sampled randomly
│   ├── 03-Modelling-SAM-distillation.ipynb           # Fine tuning SAM via training domain-specific decoder
│   ├── 02-Modelling-Active-Learning.ipynb            # Active Learning experiments with various acquisition functions
├── (sam)                                             # To contain the downloaded SAM weights
├── (src)                                             # the main source code directory
│   ├── datasets.py                                   # custom dataset classes and wrappers
│   ├── models.py                                     # custom models for segmentation
│   ├── strategies.py                                 # contains various Active Learning acquisition functions
│   ├── utils.py                                      # misclessaneous helper functions for all experiments
├── codemeta.json                                     # code metadata
├── conda_packages.txt                                # required conda packages
├── dataprep.py                                       # Python script to convert spectrograms to images and save them for use throughout
├── LICENSE.txt                                       # the full license which this project employs
├── metadata.yml                                      # project metadata
├── pip_requirements.txt                              # required pip packages
├── README.md                                         # this README file

Acknowledgement

Europlanet 2024 RI has received funding from the European Union’s Horizon 2020 research and innovation programme under grant agreement No 871149.