StarcNet: Machine Learning for Star Cluster Identification*

Pérez, Gustavo; Messa, Matteo; Calzetti, Daniela; Maji, Subhransu; Jung, Daeseong; Adamo, Angela; Sirressi, Mattia

doi:10.3847/1538-4357/abceba

Cited by 20 publications

(26 citation statements)

References 52 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…This indicates that crowding has a stronger impact than brightness on the classification of Class 1 and Class 2 objects, the main focus of this paper. Pérez et al (2021) report similar issues with Class 2 and Class 3 classifications, hence this effect appears to be inherent in the difficulty of classifying these objects in crowded regions rather than in the particular machine learning classification method employed.…”

Section: Agreement As a Function Of Crowdingmentioning

confidence: 86%

“…A citizen science approach to cluster classification was used for the PHAT (Panchromatic Hubble Andromeda Treasury) project (Johnson et al 2012;Johnson et al 2015) to accelerate classification. While this works well for nearby well-resolved clusters, exploratory efforts for the more distant galaxies in the Legacy ExtraGalactic Ultraviolet Survey - (Calzetti et al 2015 -LEGUS) project (out to ∼11 Mpc) were unsuccessful due to the more subtle differences between clusters, associations, stars, and interlopers due to the decreased physical resolution, as reported in Pérez et al (2021).…”

mentioning

confidence: 94%

“…(300 galaxies × 3000 clusters) / (10 galaxies × 300 clusters)], with a total number of clusters that could be classified around 1 million. Classification of this many objects represents a limiting constraint for the study of clusters in nearby galaxies, and was the primary reason for development of automated, neural network methods for cluster classifications (Messa et al 2018;Bialopetravičius;Narbutis & Vansevičius 2019;Grasha et al 2019;Wei et al 2020;Pérez et al 2021).…”

mentioning

confidence: 99%

“…recovery rates on the order 60 to 70 per cent between human and machine learning 1 https://archive.stsci.edu/hlsp/phangs-hst 2 https://sites.google.com/view/phangs/home approaches), exploratory efforts for the more distant galaxies, and for the compact associations, showed lower recovery rates. Subsequently, Pérez et al (2021) developed multiscale convolutional neural network models for the LEGUS project, with agreement fractions on par with the performance of our models in Wei et al (2020). An additional contribution is Bialopetravičius et al (2019), who used neural network classifications based on simulations for resolved star clusters in M31.…”

mentioning

confidence: 99%

See 3 more Smart Citations

Star cluster classification in the PHANGS–HST survey: Comparison between human and machine learning approaches

Whitmore

Lee

Chandar

et al. 2021

Monthly Notices of the Royal Astronomical Society

View full text Add to dashboard Cite

When completed, the PHANGS–HST project will provide a census of roughly 50 000 compact star clusters and associations, as well as human morphological classifications for roughly 20 000 of those objects. These large numbers motivated the development of a more objective and repeatable method to help perform source classifications. In this paper, we consider the results for five PHANGS–HST galaxies (NGC 628, NGC 1433, NGC 1566, NGC 3351, NGC 3627) using classifications from two convolutional neural network architectures (RESNET and VGG) trained using deep transfer learning techniques. The results are compared to classifications performed by humans. The primary result is that the neural network classifications are comparable in quality to the human classifications with typical agreement around 70 to 80 per cent for Class 1 clusters (symmetric, centrally concentrated) and 40 to 70 per cent for Class 2 clusters (asymmetric, centrally concentrated). If Class 1 and 2 are considered together the agreement is 82 ± 3 per cent. Dependencies on magnitudes, crowding, and background surface brightness are examined. A detailed description of the criteria and methodology used for the human classifications is included along with an examination of systematic differences between PHANGS–HST and LEGUS. The distribution of data points in a colour–colour diagram is used as a ‘figure of merit’ to further test the relative performances of the different methods. The effects on science results (e.g. determinations of mass and age functions) of using different cluster classification methods are examined and found to be minimal.

show abstract

Section: Agreement As a Function Of Crowdingmentioning

confidence: 86%

mentioning

confidence: 94%

mentioning

confidence: 99%

mentioning

confidence: 99%

See 2 more Smart Citations

Star cluster classification in the PHANGS–HST survey: Comparison between human and machine learning approaches

Whitmore

Lee

Chandar

et al. 2021

Monthly Notices of the Royal Astronomical Society

View full text Add to dashboard Cite

show abstract

“…(300 galaxies × 3,000 clusters) / (10 galaxies × 300 clusters)], with a total number of clusters that could be classified around 1 million. Classification of this many objects represents a limiting constraint for the study of clusters in nearby galaxies, and was the primary reason for development of automated, neural network methods for cluster classifications (Messa et al 2018, Grasha et al 2019, Bialopetravičius et al 2019, Wei et al 2020, Pérez et al 2021.…”

Section: Introductionmentioning

confidence: 99%

Star Cluster Classification in the PHANGS-HST Survey: Comparison between Human and Machine Learning Approaches

Whitmore,

Lee,

Chandar

et al. 2021

Preprint

View full text Add to dashboard Cite

When completed, the PHANGS-HST project will provide a census of roughly 50,000 compact star clusters and associations, as well as human morphological classifications for roughly 20,000 of those objects. These large numbers motivated the development of a more objective and repeatable method to help perform source classifications. In this paper we consider the results for five PHANGS-HST galaxies (NGC 628, NGC 1433, NGC 1566, NGC 3351, NGC 3627) using classifications from two convolutional neural network architectures (RESNET and VGG) trained using deep transfer learning techniques. The results are compared to classifications performed by humans. The primary result is that the neural network classifications are comparable in quality to the human classifications with typical agreement around 70 to 80 % for Class 1 clusters (symmetric, centrally concentrated) and 40 to 70 % for Class 2 clusters (asymmetric, centrally concentrated). If Class 1 and 2 are considered together the agreement is 82 ± 3 %. Dependencies on magnitudes, crowding, and background surface brightness are examined. A detailed description of the criteria and methodology used for the human classifications is included along with an examination of systematic differences between PHANGS-HST and LEGUS. The distribution of data points in a colour-colour diagram is used as a "figure of merit" to further test the relative performances of the different methods. The effects on science results (e.g., determinations of mass and age functions) of using different cluster classification methods are examined and found to be minimal.

show abstract

Identification of new M 31 star cluster candidates from PAndAS images using convolutional neural networks

Wang

Chen

et al. 2022

A&A

View full text Add to dashboard Cite

Context. Identification of new star cluster candidates in M 31 is fundamental for the study of the M 31 stellar cluster system. The machine-learning method convolutional neural network (CNN) is an efficient algorithm for searching for new M 31 star cluster candidates from tens of millions of images from wide-field photometric surveys. Aims. We search for new M 31 cluster candidates from the high-quality g- and i-band images of 21 245 632 sources obtained from the Pan-Andromeda Archaeological Survey (PAndAS) through a CNN. Methods. We collected confirmed M 31 clusters and noncluster objects from the literature as our training sample. Accurate double-channel CNNs were constructed and trained using the training samples. We applied the CNN classification models to the PAndAS g- and i-band images of over 21 million sources to search new M 31 cluster candidates. The CNN predictions were finally checked by five experienced human inspectors to obtain high-confidence M 31 star cluster candidates. Results. After the inspection, we identified a catalogue of 117 new M 31 cluster candidates. Most of the new candidates are young clusters that are located in the M 31 disk. Their morphology, colours, and magnitudes are similar to those of the confirmed young disk clusters. We also identified eight globular cluster candidates that are located in the M 31 halo and exhibit features similar to those of confirmed halo globular clusters. The projected distances to the M 31 centre for three of them are larger than 100 kpc.

show abstract

StarcNet: Machine Learning for Star Cluster Identification*

Cited by 20 publications

References 52 publications

Star cluster classification in the PHANGS–HST survey: Comparison between human and machine learning approaches

Star cluster classification in the PHANGS–HST survey: Comparison between human and machine learning approaches

Star Cluster Classification in the PHANGS-HST Survey: Comparison between Human and Machine Learning Approaches

Identification of new M 31 star cluster candidates from PAndAS images using convolutional neural networks

Contact Info

Product

Resources

About