---
task_categories:
- image-classification
tags:
- astrophysics
- flares
- solar flares
- sun
pretty_name: e-Callisto Solar Flare Detection
size_categories:
- 100K<n<1M
---

# e-Callisto Solar Flare Detection Dataset

![](https://www.fhnw.ch/en/++theme++web16theme/assets/media/img/university-applied-sciences-arts-northwestern-switzerland-fhnw-logo.svg)
[Institute of Data Science i4Ds, FHNW](https://i4ds.ch)

## Overview
This dataset comprises radio spectra from the [e-Callisto solar spectrometer network](https://www.e-callisto.org/index.html), annotated based on [labels from the e-Callisto database](http://soleil.i4ds.ch/solarradio/data/BurstLists/2010-yyyy_Monstein/). 
The data was downloaded using the [ecallisto_ng Package](https://github.com/i4Ds/ecallisto_ng). It's designed for training machine learning models to automatically detect and classify solar flares.

## Data Collection
Data has been collected from various stations, with the following date ranges:

| Station           | Date Range               |
|-------------------|--------------------------|
| Australia-ASSA_01 | 2021-02-13 to 2021-12-11 |
| Australia-ASSA_02 | 2021-02-13 to 2021-12-09 |
| Australia-ASSA_62 | 2021-12-10 to 2023-12-12 |
| Australia-ASSA_63 | 2021-12-10 to 2023-12-12 |

## Data Augmentation
Due to the rarity of solar flares, we've augmented the dataset by padding the time series data around each flare event.

## Caution
The dataset underwent preprocessing and certain assumptions were made for label cleanup. Be aware of potential inaccuracies in the labels.

## Split Recommendations
The dataset doesn't include predefined train-validation-test splits. When creating splits, ensure augmented data does not overlap between training and validation/test sets to avoid data leakage.