Published October 30, 2021 | Version v1

Whistler segmentation training dataset

  • 1. Department of Physics of Complex Systems, ELTE Eötvös Loránd University, Budapest, Hungary
  • 2. Space Research Group, Department of Geophysics and Space Sciences, ELTE Eötvös Loránd University, Budapest, Hungary
  • 3. British Antarctic Survey, Cambridge, UK.

Description

Generated whistlers with pixel-segmented separate masks for each trace.

File naming:
-  train_images.tar.gz contains all the train images (>10 GB)
- example_images.zip contains 5 spectrograms and the corresponding segmentation masks. (~5 MB))
- {id}_rothera_{n}_y.png, {id}_rothera_x.png, where {id} is a unique ID to match the masks and the spectrogram image, {n} enumerates the whistler traces within a spectrogram. Files ending _y.png are the bitmask images, while the _x.png files are the spectrograms.

Opening the bitmasks in a regular image viewer might show only a black image. In this case, opening as a matrix (eg via matplotlib or opencv in python), converting to binary values, and visualizing them can help.

To be updated upon acceptance of our research paper.

Files

example_images.zip

Files (10.7 GB)

Name Size
md5:50b81803a605ce2c615e7ee6eea43273
5.9 MB Preview Download
md5:94e85ddea2cccbb81f5e47f14b9c7edc
10.7 GB Download