Home > Research > Publications & Outputs > Transfer learning for galaxy feature detection

Links

Text available via DOI:

View graph of relations

Transfer learning for galaxy feature detection: Finding giant star-forming clumps in low-redshift galaxies using Faster Region-based Convolutional Neural Network

Research output: Contribution to Journal/MagazineJournal articlepeer-review

Published

Standard

Transfer learning for galaxy feature detection: Finding giant star-forming clumps in low-redshift galaxies using Faster Region-based Convolutional Neural Network. / Popp, Jürgen J; Dickinson, Hugh; Serjeant, Stephen et al.
In: RAS Techniques and Instruments, Vol. 3, No. 1, 31.12.2024, p. 174-197.

Research output: Contribution to Journal/MagazineJournal articlepeer-review

Harvard

Popp, JJ, Dickinson, H, Serjeant, S, Walmsley, M, Adams, D, Fortson, L, Mantha, K, Mehta, V, Dawson, JM, Kruk, S & Simmons, B 2024, 'Transfer learning for galaxy feature detection: Finding giant star-forming clumps in low-redshift galaxies using Faster Region-based Convolutional Neural Network', RAS Techniques and Instruments, vol. 3, no. 1, pp. 174-197. https://doi.org/10.1093/rasti/rzae013

APA

Popp, J. J., Dickinson, H., Serjeant, S., Walmsley, M., Adams, D., Fortson, L., Mantha, K., Mehta, V., Dawson, J. M., Kruk, S., & Simmons, B. (2024). Transfer learning for galaxy feature detection: Finding giant star-forming clumps in low-redshift galaxies using Faster Region-based Convolutional Neural Network. RAS Techniques and Instruments, 3(1), 174-197. https://doi.org/10.1093/rasti/rzae013

Vancouver

Popp JJ, Dickinson H, Serjeant S, Walmsley M, Adams D, Fortson L et al. Transfer learning for galaxy feature detection: Finding giant star-forming clumps in low-redshift galaxies using Faster Region-based Convolutional Neural Network. RAS Techniques and Instruments. 2024 Dec 31;3(1):174-197. Epub 2024 Apr 17. doi: 10.1093/rasti/rzae013

Author

Popp, Jürgen J ; Dickinson, Hugh ; Serjeant, Stephen et al. / Transfer learning for galaxy feature detection : Finding giant star-forming clumps in low-redshift galaxies using Faster Region-based Convolutional Neural Network. In: RAS Techniques and Instruments. 2024 ; Vol. 3, No. 1. pp. 174-197.

Bibtex

@article{bd5b20ce374a48f8aa48b5c1f8d885b7,
title = "Transfer learning for galaxy feature detection: Finding giant star-forming clumps in low-redshift galaxies using Faster Region-based Convolutional Neural Network",
abstract = "Giant star-forming clumps (GSFCs) are areas of intensive star-formation that are commonly observed in high-redshift (z ≳ 1) galaxies but their formation and role in galaxy evolution remain unclear. Observations of low-redshift clumpy galaxy analogues are rare but the availability of wide-field galaxy survey data makes the detection of large clumpy galaxy samples much more feasible. Deep Learning (DL), and in particular Convolutional Neural Networks (CNNs), have been successfully applied to image classification tasks in astrophysical data analysis. However, one application of DL that remains relatively unexplored is that of automatically identifying and localizing specific objects or features in astrophysical imaging data. In this paper, we demonstrate the use of DL-based object detection models to localize GSFCs in astrophysical imaging data. We apply the Faster Region-based Convolutional Neural Network object detection framework (FRCNN) to identify GSFCs in low-redshift (z ≲ 0.3) galaxies. Unlike other studies, we train different FRCNN models on observational data that was collected by the Sloan Digital Sky Survey and labelled by volunteers from the citizen science project {\textquoteleft}Galaxy Zoo: Clump Scout{\textquoteright}. The FRCNN model relies on a CNN component as a {\textquoteleft}backbone{\textquoteright} feature extractor. We show that CNNs, that have been pre-trained for image classification using astrophysical images, outperform those that have been pre-trained on terrestrial images. In particular, we compare a domain-specific CNN – {\textquoteleft}Zoobot{\textquoteright} – with a generic classification backbone and find that Zoobot achieves higher detection performance. Our final model is capable of producing GSFC detections with a completeness and purity of ≥0.8 while only being trained on ∼5000 galaxy images.",
keywords = "Object Detection, Deep Learning, Machine Learning, Transfer Learning, Galaxies: Structure, Data Methods",
author = "Popp, {J{\"u}rgen J} and Hugh Dickinson and Stephen Serjeant and Mike Walmsley and Dominic Adams and Lucy Fortson and Kameswara Mantha and Vihang Mehta and Dawson, {James M} and Sandor Kruk and Brooke Simmons",
year = "2024",
month = dec,
day = "31",
doi = "10.1093/rasti/rzae013",
language = "English",
volume = "3",
pages = "174--197",
journal = "RAS Techniques and Instruments",
issn = "2752-8200",
publisher = "Oxford University Press (OUP)",
number = "1",

}

RIS

TY - JOUR

T1 - Transfer learning for galaxy feature detection

T2 - Finding giant star-forming clumps in low-redshift galaxies using Faster Region-based Convolutional Neural Network

AU - Popp, Jürgen J

AU - Dickinson, Hugh

AU - Serjeant, Stephen

AU - Walmsley, Mike

AU - Adams, Dominic

AU - Fortson, Lucy

AU - Mantha, Kameswara

AU - Mehta, Vihang

AU - Dawson, James M

AU - Kruk, Sandor

AU - Simmons, Brooke

PY - 2024/12/31

Y1 - 2024/12/31

N2 - Giant star-forming clumps (GSFCs) are areas of intensive star-formation that are commonly observed in high-redshift (z ≳ 1) galaxies but their formation and role in galaxy evolution remain unclear. Observations of low-redshift clumpy galaxy analogues are rare but the availability of wide-field galaxy survey data makes the detection of large clumpy galaxy samples much more feasible. Deep Learning (DL), and in particular Convolutional Neural Networks (CNNs), have been successfully applied to image classification tasks in astrophysical data analysis. However, one application of DL that remains relatively unexplored is that of automatically identifying and localizing specific objects or features in astrophysical imaging data. In this paper, we demonstrate the use of DL-based object detection models to localize GSFCs in astrophysical imaging data. We apply the Faster Region-based Convolutional Neural Network object detection framework (FRCNN) to identify GSFCs in low-redshift (z ≲ 0.3) galaxies. Unlike other studies, we train different FRCNN models on observational data that was collected by the Sloan Digital Sky Survey and labelled by volunteers from the citizen science project ‘Galaxy Zoo: Clump Scout’. The FRCNN model relies on a CNN component as a ‘backbone’ feature extractor. We show that CNNs, that have been pre-trained for image classification using astrophysical images, outperform those that have been pre-trained on terrestrial images. In particular, we compare a domain-specific CNN – ‘Zoobot’ – with a generic classification backbone and find that Zoobot achieves higher detection performance. Our final model is capable of producing GSFC detections with a completeness and purity of ≥0.8 while only being trained on ∼5000 galaxy images.

AB - Giant star-forming clumps (GSFCs) are areas of intensive star-formation that are commonly observed in high-redshift (z ≳ 1) galaxies but their formation and role in galaxy evolution remain unclear. Observations of low-redshift clumpy galaxy analogues are rare but the availability of wide-field galaxy survey data makes the detection of large clumpy galaxy samples much more feasible. Deep Learning (DL), and in particular Convolutional Neural Networks (CNNs), have been successfully applied to image classification tasks in astrophysical data analysis. However, one application of DL that remains relatively unexplored is that of automatically identifying and localizing specific objects or features in astrophysical imaging data. In this paper, we demonstrate the use of DL-based object detection models to localize GSFCs in astrophysical imaging data. We apply the Faster Region-based Convolutional Neural Network object detection framework (FRCNN) to identify GSFCs in low-redshift (z ≲ 0.3) galaxies. Unlike other studies, we train different FRCNN models on observational data that was collected by the Sloan Digital Sky Survey and labelled by volunteers from the citizen science project ‘Galaxy Zoo: Clump Scout’. The FRCNN model relies on a CNN component as a ‘backbone’ feature extractor. We show that CNNs, that have been pre-trained for image classification using astrophysical images, outperform those that have been pre-trained on terrestrial images. In particular, we compare a domain-specific CNN – ‘Zoobot’ – with a generic classification backbone and find that Zoobot achieves higher detection performance. Our final model is capable of producing GSFC detections with a completeness and purity of ≥0.8 while only being trained on ∼5000 galaxy images.

KW - Object Detection

KW - Deep Learning

KW - Machine Learning

KW - Transfer Learning

KW - Galaxies: Structure

KW - Data Methods

U2 - 10.1093/rasti/rzae013

DO - 10.1093/rasti/rzae013

M3 - Journal article

VL - 3

SP - 174

EP - 197

JO - RAS Techniques and Instruments

JF - RAS Techniques and Instruments

SN - 2752-8200

IS - 1

ER -