astrods provides a collection of ready-to-use datasets in Astronomy for use with TensorFlow, Jax, and other Machine Learning frameworks. The datasets have been built with the tensorflow-datasets (tfds) dataset builder and therefore it handles the downloading and preparing of the data deterministically and the construction of the tf.data.Dataset object.
The current datasets available are:
| Name | reference | size | splits |
|---|---|---|---|
| XCLASS | Kosiba et al. (2020) | 259.3 MB | ('train', 'test') |
| galaxyMNIST | Walmsley et al (2022) | 1.13 GB | ('train', 'test') |
astrods requires tensorflow and tensorflow-datasets to be installed pip install tensorflow-datasets
Import libraries and download data
import astrods.XCLASS #replace XCLASS with dataset of interest
import tensorflow_datasets as tfds
ds, info = tfds.load('XCLASS', split='train', shuffle_files=True, with_info=True) #replace XCLASS with dataset of interest
print(info)
Example datasample
sample = ds.take(1)
for ex in sample:
print(ex)