Tidalsaurus Datasets and Models

This contains the datasets and pre-trained models described in Desmons et al. (2024). This work focuses on identifying galaxy tidal features using self-supervised machine learning.

The dataset folder includes two Tensorflow datasets, an unlabelled dataset with ~45,000 objects "Unlabelled_ds", and a tidal feature dataset with ~400 objects "Tidal_ds". Each dataset consists of 128x128 pixel postage stamps of HSC-SSP galaxies, corresponding HSC-SSP object IDs, and galaxy attributes.

Also included are two pre-trained model designed to be loaded using the tensorflow.keras.models.load_model() command. The pre-trained models include the self-supervised nearest-neighbour model "Trained_NNCLR_model" and the supervised finetuned classifier "Trained_finetuned_classifier".

The details of the project, the python code, and the instructions on using the code to reproduce the results presented in the paper can be found at https://github.com/LSSTISSC/Tidalsaurus