pith. machine review for the scientific record. sign in

arxiv: 2604.10086 · v1 · submitted 2026-04-11 · 🌌 astro-ph.IM

Recognition: unknown

A Multi-modal Fusion Network for Star-Galaxy Classification from CSST Simulated Datasets

Zhuoming Han , Tianmeng Zhang , Chao Liu , Chenxiaoji Ling
Authors on Pith no claims yet

Pith reviewed 2026-05-10 15:52 UTC · model grok-4.3

classification 🌌 astro-ph.IM
keywords star-galaxy classificationmulti-modal fusiondeep learningResNet-50BiLSTMCSST simulationsastronomical object classificationcatalog and image data
0
0 comments X

The pith

A multi-modal network fuses CSST image and catalog data to classify stars and galaxies with recalls above 99.6 percent.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper builds a dataset of simulated CSST images paired with photometric catalogs containing over 125,000 stars and galaxies. It then trains a supervised network that extracts visual features from the images and sequential features from the catalog measurements before combining them for the final decision. This matters because next-generation surveys will produce petabytes of data where manual or catalog-only separation becomes impractical. The resulting model reaches 99.81 percent recall for galaxies and 99.66 percent recall for stars after 50 training epochs. The authors also show that the same network retains strong performance on faint objects and high-redshift galaxies.

Core claim

The central claim is that a supervised network combining ResNet-50 image feature extraction with BiLSTM catalog feature processing, followed by multi-modal fusion, produces star-galaxy classifications with 99.81 percent galaxy recall and 99.66 percent star recall on a CSST simulation dataset of 32,371 stars and 93,525 galaxies.

What carries the argument

The ResNet-50 and BiLSTM fusion network that merges spatial image features with photometric catalog sequences for binary classification.

If this is right

  • The model continues to deliver high accuracy on faint astronomical objects.
  • Performance stays strong for high-redshift galaxies.
  • Data augmentation and multi-modal fusion each improve results over single-modality baselines.
  • The approach is positioned for direct use on the large volumes of data expected from the CSST survey.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • If the simulation fidelity holds, the same architecture could be retrained or fine-tuned on early real CSST releases without major redesign.
  • The fusion method may generalize to other multi-band surveys where both imaging and photometry are available.
  • Direct head-to-head tests against purely image-based or purely catalog-based classifiers on identical CSST data would quantify the exact contribution of the fusion step.

Load-bearing premise

The CSST simulated images and catalogs accurately reproduce the noise, point-spread functions, and source distributions of actual telescope observations.

What would settle it

Apply the trained model to a sample of real CSST observations and measure whether recall for stars and galaxies remains above 99 percent or drops substantially.

Figures

Figures reproduced from arXiv: 2604.10086 by Chao Liu, Chenxiaoji Ling, Tianmeng Zhang, Zhuoming Han.

Figure 1
Figure 1. Figure 1: The sky region coverage of simulation data. The blue squares represent single [PITH_FULL_IMAGE:figures/full_fig_p006_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Cutout image for 7 bands (from left to right: [PITH_FULL_IMAGE:figures/full_fig_p008_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: The r-band magnitude distribution of stars (red) and galaxies (blue) from pho [PITH_FULL_IMAGE:figures/full_fig_p009_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: ResNet-50, which consists of five stages, is employed as the backbone [PITH_FULL_IMAGE:figures/full_fig_p010_4.png] view at source ↗
Figure 4
Figure 4. Figure 4: RBiM network. The 7-channel image data and catalog are processed through [PITH_FULL_IMAGE:figures/full_fig_p011_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: Residual connection in Stage 1 of ResNet-50, including two types of residual [PITH_FULL_IMAGE:figures/full_fig_p012_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: Confusion matrix for classification of stars and galaxies. The vertical axis [PITH_FULL_IMAGE:figures/full_fig_p014_6.png] view at source ↗
Figure 7
Figure 7. Figure 7: Confusion matrix for the data missing red or blue bands. Left figure: Confusion [PITH_FULL_IMAGE:figures/full_fig_p017_7.png] view at source ↗
Figure 8
Figure 8. Figure 8: The variation of spread_model with respect to kron_mag. The horizontal axis represents the NUV band magnitude of a brick, while the vertical axis shows the spread_model parameter calculated by SExtractor. Star and galaxy samples are repre￾sented by red and blue points, respectively. The legend indicates the source density, with darker areas representing a greater number of sources. the star. The error rate… view at source ↗
Figure 9
Figure 9. Figure 9: The comparison of the classification effect of the network and spread model [PITH_FULL_IMAGE:figures/full_fig_p019_9.png] view at source ↗
Figure 10
Figure 10. Figure 10: The comparison of the classification effect of the network and spread model [PITH_FULL_IMAGE:figures/full_fig_p020_10.png] view at source ↗
read the original abstract

The distinction between stars and galaxies is a fundamental problem in the field of celestial classification. This issue has become challenging for these ongoing and upcoming digital surveys, which will produce terabytes and even petabytes of astronomical data. While deep learning offers a powerful solution for star-galaxy classification in large-scale datasets, most current approaches are limited by their reliance on catalog data alone, which consists primarily of multi-band magnitudes and imprecise morphological parameters. Therefore, we utilize China Space Station Telescope (CSST) simulation data to build a dataset with both image and photometric catalog, including 32,371 stars and 93,525 galaxies. A supervised deep learning network based on ResNet-50 and BiLSTM is proposed to improve the classification of two types of astronomical objects. The features of the catalog and image are integrated by the model, achieving 99.81% recall for galaxies and 99.66% recall for stars after training on GPU for 50 epochs. We evaluated the effects of data augmentation and multi-modal data fusion, which demonstrate that our model has commendable performance. Furthermore, our model also has a high accuracy rate for faint astronomical objects and high redshift galaxies, demonstrating its applicability to the upcoming CSST scientific data.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

4 major / 2 minor

Summary. The paper proposes a supervised multi-modal deep learning model that fuses ResNet-50 image features with BiLSTM catalog features for star-galaxy classification on CSST simulated data (32k stars, 93k galaxies). It reports recalls of 99.81% (galaxies) and 99.66% (stars) on a held-out test set after 50 GPU epochs, evaluates data augmentation and fusion, and asserts strong performance on faint objects and high-redshift galaxies with applicability to real CSST observations.

Significance. A working multi-modal fusion approach could aid efficient classification in petabyte-scale surveys by combining imaging and photometry; the reported recalls on simulated data are high, but the result's significance for real CSST data is constrained by the absence of any demonstrated fidelity between the simulations and actual telescope noise, PSF, or distributions.

major comments (4)
  1. [Abstract] Abstract and final paragraph: the claim that the model demonstrates 'applicability to the upcoming CSST scientific data' rests on the untested premise that the simulated catalog and images reproduce the joint distributions of magnitudes, colors, sizes, redshifts, noise, and PSF that real CSST observations will exhibit; no KS tests, moment matching, or residual comparisons to any real reference catalog or image set are supplied.
  2. [Results] Results section: the statements of 'high accuracy rate for faint astronomical objects and high redshift galaxies' are presented without supporting per-bin statistics, confusion matrices, or performance curves as a function of magnitude or redshift, leaving the claim without quantitative backing.
  3. [Methods/Results] Methods and Results: no baseline comparisons (traditional color-color or morphology cuts, single-modal ResNet or catalog-only networks) or ablation tables quantifying the gain from multi-modal fusion are reported, so the improvement attributable to the ResNet-50 + BiLSTM architecture cannot be assessed.
  4. [Abstract] Abstract and Methods: the headline recalls lack error bars, cross-validation details, or indications of whether the held-out test set was used in any hyperparameter selection, making it impossible to judge whether the 99.81 % / 99.66 % figures are robust or over-optimistic.
minor comments (2)
  1. [Methods] The training/validation/test split ratios and any class-balancing strategy are not stated, although the raw counts (32,371 stars, 93,525 galaxies) are given.
  2. Figure captions and axis labels should explicitly indicate whether the displayed metrics are on the training, validation, or test set.

Simulated Author's Rebuttal

4 responses · 0 unresolved

We thank the referee for the constructive and detailed comments. We have addressed each major point below and revised the manuscript accordingly where feasible to improve clarity and rigor.

read point-by-point responses

  1. Referee: [Abstract] Abstract and final paragraph: the claim that the model demonstrates 'applicability to the upcoming CSST scientific data' rests on the untested premise that the simulated catalog and images reproduce the joint distributions of magnitudes, colors, sizes, redshifts, noise, and PSF that real CSST observations will exhibit; no KS tests, moment matching, or residual comparisons to any real reference catalog or image set are supplied.

    Authors: We agree that direct statistical comparisons to real CSST data are not possible, as the telescope has not yet begun operations. The simulations were constructed using CSST-specific instrument models for noise, PSF, and source distributions, but we acknowledge this leaves the real-world fidelity untested. In the revised manuscript we will tone down the applicability language in the abstract and conclusion to 'potential applicability to upcoming CSST data' and add an explicit limitations paragraph discussing simulation assumptions. revision: partial

  2. Referee: [Results] Results section: the statements of 'high accuracy rate for faint astronomical objects and high redshift galaxies' are presented without supporting per-bin statistics, confusion matrices, or performance curves as a function of magnitude or redshift, leaving the claim without quantitative backing.

    Authors: We accept that the current text lacks the requested quantitative support. The revised Results section will include new figures and tables reporting recall and precision binned by magnitude and redshift, together with confusion matrices for the faint and high-redshift subsets. revision: yes

  3. Referee: [Methods/Results] Methods and Results: no baseline comparisons (traditional color-color or morphology cuts, single-modal ResNet or catalog-only networks) or ablation tables quantifying the gain from multi-modal fusion are reported, so the improvement attributable to the ResNet-50 + BiLSTM architecture cannot be assessed.

    Authors: While the manuscript already evaluates the contribution of multi-modal fusion versus single-modality inputs, we agree that explicit baselines against classical methods are missing. We will add a dedicated comparison subsection that includes color-color and morphology-cut baselines as well as single-modal ResNet-50 and BiLSTM results, with an ablation table quantifying the fusion gain. revision: yes

  4. Referee: [Abstract] Abstract and Methods: the headline recalls lack error bars, cross-validation details, or indications of whether the held-out test set was used in any hyperparameter selection, making it impossible to judge whether the 99.81 % / 99.66 % figures are robust or over-optimistic.

    Authors: We will strengthen the statistical reporting. The revised abstract and Methods section will report error bars obtained from multiple independent training runs, describe the train/validation/test split, and clarify that hyperparameter choices were made on the validation set only. We will also add a brief k-fold cross-validation experiment if space permits. revision: yes

Circularity Check

0 steps flagged

No circularity; standard supervised training and held-out evaluation on simulated data

full rationale

The paper trains a ResNet-50 + BiLSTM fusion network on a fixed CSST simulation dataset (32k stars, 93k galaxies) and reports recall on a held-out test split after 50 epochs. No mathematical derivations, equations, or first-principles predictions appear in the provided text. Performance numbers are direct empirical measurements, not quantities that reduce to fitted parameters or self-citations by construction. The inference to real CSST observations rests on an untested simulation-fidelity assumption, but this is a correctness risk, not a circular reduction within the derivation chain. The work is self-contained as a standard ML classification experiment.

Axiom & Free-Parameter Ledger

1 free parameters · 1 axioms · 0 invented entities

The central claim rests on standard supervised deep-learning assumptions plus the unverified premise that the provided simulation faithfully represents future real observations.

free parameters (1)
  • Training hyperparameters
    Learning rate, batch size, optimizer, and exact fusion-layer dimensions are not reported but must have been chosen to reach the stated performance.
axioms (1)
  • domain assumption Simulated CSST data distribution matches real telescope observations
    The model is trained and evaluated exclusively on simulation; transfer to real data is asserted without quantitative evidence in the abstract.

pith-pipeline@v0.9.0 · 5522 in / 1360 out tokens · 49638 ms · 2026-05-10T15:52:46.595869+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

42 extracted references · 36 canonical work pages · 2 internal anchors

  1. [1]

    write newline

    " write newline "" before.all 'output.state := FUNCTION n.dashify 't := "" t empty not t #1 #1 substring "-" = t #1 #2 substring "--" = not "--" * t #2 global.max substring 't := t #1 #1 substring "-" = "-" * t #2 global.max substring 't := while if t #1 #1 substring * t #2 global.max substring 't := if while FUNCTION word.in bbl.in ":" * " " * FUNCTION f...

  2. [2]

    Dark Energy Survey Collaboration, The Dark Energy Survey: more than dark energy - an overview, Mon. Not. R. Astron. Soc., 460 (2) (2016), pp. 1270-1299, doi: 10.1093/mnras/stw641

    work page doi:10.1093/mnras/stw641 2016

  3. [3]

    Donald G. York, J. Adelman, John E. Anderson, et al., The Sloan Digital Sky Survey: Technical Summary, Astron. J., 120 (3) (2000), pp. 1579-1587, doi: 10.1086/301513

    work page doi:10.1086/301513 2000

  4. [4]

    Wide-Field InfrarRed Survey Telescope-Astrophysics Focused Telescope Assets WFIRST-AFTA 2015 Report

    D. Spergel, N. Gehrels, C. Baltay, et al., Wide-Field Infrared Survey Telescope-Astrophysics Focused Telescope Assets WFIRST-AFTA 2015 Report, (2015), doi: 10.48550/arXiv.1503.03757

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.48550/arxiv.1503.03757 2015

  5. [5]

    Euclid Collaboration, Euclid preparation. I. The Euclid Wide Survey, Astron. Astrophys., 662 (2022), p. A112, doi: 10.1051/0004-6361/202141938

    work page doi:10.1051/0004-6361/202141938 2022

  6. [6]

    doi:10.48550/arXiv.1410.8185

    Eric C. Bellm, The Zwicky Transient Facility, The Third Hot-wiring the Transient Universe Workshop, (2014), pp. 27-33, doi: 10.48550/arXiv.1410.8185

    work page doi:10.48550/arxiv.1410.8185 2014

  7. [7]

    K. C. Chambers, E. A. Magnier, N. Metcalfe, et al., The Pan-STARRS1 Surveys, (2016), doi: 10.48550/arXiv.1612.05560

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.48550/arxiv.1612.05560 2016

  8. [8]

    Željko Ivezić, Steven M. Kahn, J. Anthony Tyson, et al., Lsst: from science drivers to reference design and anticipated data products, Astrophys. J., 873 (2) (2019), p. 111, doi: 10.3847/1538-4357/ab042c

    work page doi:10.3847/1538-4357/ab042c 2019

  9. [9]

    2016 b , , 595, A2, 10.1051/0004-6361/201629512

    Gaia Collaboration, Gaia Data Release 1. Summary of the astrometric, photometric, and survey properties, Astron. Astrophys., 595 (A2) (2016), p. 23, doi: 10.1051/0004-6361/201629512

    work page doi:10.1051/0004-6361/201629512 2016

  10. [10]

    E1.16-4-18

    Zhan, Hu, An Overview of the Chinese Space Station Optical Survey, 42nd COSPAR Scientific Assembly, (2018), Abstract id. E1.16-4-18

    2018

  11. [11]

    Fatemeh Zahra Zeraatgari, Fatemeh Hafezianzade, Yanxia Zhang, et al., Machine learning-based photometric classification of galaxies, quasars, emission-line galaxies, and stars, Mon. Not. R. Astron. Soc., 527 (3) (2024), pp. 4677–4689, doi: 10.1093/mnras/stad3436

    work page doi:10.1093/mnras/stad3436 2024

  12. [12]

    R Moradi, F Rastegarnia, Y Wang, & M T Mirtorabi, FNet II: Spectral Classification of Quasars, Galaxies, Stars, and broad absorption line (BAL) Quasars, Mon. Not. R. Astron. Soc., 533 (2) (2024), pp. 1976–1985, doi: 10.1093/mnras/stae1878

    work page doi:10.1093/mnras/stae1878 2024

  13. [13]

    , keywords =

    Kron, R. G., Photometry of a complete sample of faint galaxies, Astrophys. J. Suppl. Ser., 43 (1980), pp. 305-325, doi: 10.1086/190669

    work page doi:10.1086/190669 1980

  14. [14]

    S. C. Odewahn, R. R. de Carvalho, R. R. Gal, et al., The Digitized Second Palomar Observatory Sky Survey (DPOSS). III. Star-Galaxy Separation, Astron. J., 128 (6) (2004), pp. 3092-3107, doi: 10.1086/425525

    work page doi:10.1086/425525 2004

  15. [15]

    E. C. Vasconcellos, R. R. de Carvalho, R. R. Gal, et al., Decision Tree Classifiers for Star/Galaxy Separation, Astron. J., 141 (6) (2011), p. 189, doi: 10.1088/0004-6256/141/6/189

    work page doi:10.1088/0004-6256/141/6/189 2011

  16. [16]

    Newman, Michael C

    Jeffrey A. Newman, Michael C. Cooper, Marc Davis, et al., The DEEP2 Galaxy Redshift Survey: Design, Observations, Data Reduction, and Redshifts, Astrophys. J. Suppl. Ser., 208 (1) (2013), p. 5, doi: 10.1088/0067-0049/208/1/5

    work page doi:10.1088/0067-0049/208/1/5 2013

  17. [17]

    Huang, L

    J.-S. Huang, L. L. Cowie, J. P. Gardner, et al., The Hawaii K-Band Galaxy Survey. II. Bright K-Band Imaging, Astrophys. J., 476 (1) (1997), pp. 12-21, doi: 10.1086/303598

    work page doi:10.1086/303598 1997

  18. [18]

    I. K. Baldry, A. S. G. Robotham, D. T. Hill, et al., Galaxy And Mass Assembly (GAMA): the input catalogue and star-galaxy separation, Mon. Not. R. Astron. Soc., 404 (1) (2010), pp. 86-100, doi: 10.1111/j.1365-2966.2010.16282.x

    work page doi:10.1111/j.1365-2966.2010.16282.x 2010

  19. [19]

    Małek, A

    K. Małek, A. Solarz1, A. Pollo, et al., The VIMOS Public Extragalactic Redshift Survey (VIPERS). A support vector machine classification of galaxies, stars, and AGNs, Astron. Astrophys., 557 (2013), p. A16, doi: 10.1051/0004-6361/201321447

    work page doi:10.1051/0004-6361/201321447 2013

  20. [20]

    Molino, N

    A. Molino, N. Benítez, M. Moles, et al., The ALHAMBRA Survey: Bayesian photometric redshifts with 23 bands for 3 deg, Mon. Not. R. Astron. Soc., 441 (4) (2014), pp. 2891–2922, doi: 10.1093/mnras/stu387

    work page doi:10.1093/mnras/stu387 2014

  21. [21]

    M. T. Soumagnac, F. B. Abdalla, O. Lahav, et al., Star/galaxy separation at faint magnitudes: application to a simulated Dark Energy Survey, Mon. Not. R. Astron. Soc., 450 (1) (2015), pp. 666–680, doi: 10.1093/mnras/stu1410

    work page doi:10.1093/mnras/stu1410 2015

  22. [22]

    Kim, & Robert J

    Edward J. Kim, & Robert J. Brunner, Star-galaxy classification using deep convolutional neural networks, Mon. Not. R. Astron. Soc., 464 (4) (2017), pp. 4463–4475, doi: 10.1093/mnras/stw2672

    work page doi:10.1093/mnras/stw2672 2017

  23. [23]

    Shawhan, Patrick R

    Peter S. Shawhan, Patrick R. Brady, Adam Brazier, et al., Data Analysis Challenges for Multi-Messenger Astrophysics, ASP Conf.Ser., 523 (2019), p. 705

    2019

  24. [24]

    Nakazono, C

    L. Nakazono, C. Mendes de Oliveira, N. S. T. Hirata, et al., On the discovery of stars, quasars, and galaxies in the Southern Hemisphere with S-PLUS DR2, Mon. Not. R. Astron. Soc., 507 (4) (2021), pp. 5847-5868, doi: 10.1093/mnras/stab1835

    work page doi:10.1093/mnras/stab1835 2021

  25. [25]

    Stoppa, S

    F. Stoppa, S. Bhattacharyya, R. Ruiz de Austri, et al., AutoSourceID-Classifier Star-Galaxy Classification using a Convolutional Neural Network with Spatial Information, Astron. Astrophys., 680 (2023), p. A109, doi: 10.1051/0004-6361/202347576

    work page doi:10.1051/0004-6361/202347576 2023

  26. [26]

    Channappayya, P.K

    Srinadh Reddy Bhavanam, Sumohana S. Channappayya, P.K. Srijith, & Shantanu Desai, Enhanced Astronomical Source Classification with Integration of Attention Mechanisms and Vision Transformers, Astrophys Space Sci, 369 (8) (2024), p. 92, doi: 10.1007/s10509-024-04357-9

    work page doi:10.1007/s10509-024-04357-9 2024

  27. [27]

    Shiliang Zhang, Guanwen Fang, Jie Song, et al., Preparation for CSST: Star-Galaxy Classification Using a Rotationally Invariant Supervised Machine Learning Method, Res. Astron. Astrophys., 24 (9) (2024), Article 095012, doi: 10.1088/1674-4527/ad6fe6

    work page doi:10.1088/1674-4527/ad6fe6 2024

  28. [28]

    J, 165 (2) (2023), p

    GuanWen Fang, Shuo Ba, Yizhou Gu, et al., Automatic Classification of Galaxy Morphology: A Rotationally-invariant Supervised Machine-learning Method Based on the Unsupervised Machine-learning Data Set, Astron. J, 165 (2) (2023), p. 35, doi: 10.3847/1538-3881/aca1a6

    work page doi:10.3847/1538-3881/aca1a6 2023

  29. [29]

    M. V. Costa-Duarte, L. Sampedro, A. Molino, et al., The S-PLUS: A Star/Galaxy Classification Based on a Machine Learning Approach, (2019), doi: 10.48550/arXiv.1909.08626

    work page doi:10.48550/arxiv.1909.08626 2019

  30. [30]

    Cheng-Liang Wei, Guo-Liang Li, Yue-Dong Fang, et al., Mock Observations for the CSST Mission: Main Surveys-An Overview of Framework and Simulation Suite, Res. Astron. Astrophys., 26 (2) (2026), Article 024001, doi: 10.1088/1674-4527/ae20fe

    work page doi:10.1088/1674-4527/ae20fe 2026

  31. [31]

    Cheng-Liang Wei, Yu Luo, Hao Tian, et al., Mock Observations for the CSST Mission: Main Surveys–The Mock Catalog, Res. Astron. Astrophys., 26 (2) (2026), Article 024004, doi: 10.1088/1674-4527/ae20ff

    work page doi:10.1088/1674-4527/ae20ff 2026

  32. [32]

    Khoshgoftaar, A survey on Image Data Augmentation for Deep Learning, J Big Data, 6 (2019), pp

    Connor Shorten, & T. Khoshgoftaar, A survey on Image Data Augmentation for Deep Learning, J Big Data, 6 (2019), pp. 1-48, doi: 10.1186/s40537-019-0197-0

    work page doi:10.1186/s40537-019-0197-0 2019

  33. [33]

    391–405, doi: 10.1007/978-981-99-2100-3\_31

    M Sharma, P Kushwaha, P Kumari, et al., Machine Learning Techniques in Data Fusion: A Review, International Conference on Communication and Intelligent Systems, 686 (2023), pp. 391–405, doi: 10.1007/978-981-99-2100-3\_31

    work page doi:10.1007/978-981-99-2100-3 2023

  34. [34]

    Siddharth Chaini, Atharva Bagul, Anish Deshpande, et al., Photometric Identification of Compact Galaxies, Stars, and Quasars Using Multiple Neural Networks, Mon. Not. R. Astron. Soc., 518 (2) (2023), pp. 3123-3136, doi: 10.1093/mnras/stac3336

    work page doi:10.1093/mnras/stac3336 2023

  35. [35]

    Napolitano, et al., Morpho-Photometric Classification of KiDS DR5 Sources Based on Neural Networks: A Comprehensive Star-Quasar-Galaxy Catalog, Astrophys

    Hai-Cheng Feng, Rui Li, Nicola R. Napolitano, et al., Morpho-Photometric Classification of KiDS DR5 Sources Based on Neural Networks: A Comprehensive Star-Quasar-Galaxy Catalog, Astrophys. J. Suppl. Ser., 279 (1) (2025), doi: 10.3847/1538-4365/adde5a

    work page doi:10.3847/1538-4365/adde5a 2025

  36. [36]

    Adam Paszke, Sam Gross, Francisco Massa, et al., PyTorch: An Imperative Style, High-Performance Deep Learning Library, NeurIPS, 32 (2019)

    2019

  37. [37]

    Deep Residual Learning for Image Recognition , isbn =

    Kaiming He, Xiangyu Zhang, Shaoqing Ren, & Jian Sun. Deep Residual Learning for Image Recognition, IEEE, CVPR, (2016), pp. 770-778, doi: 10.1109/CVPR.2016.90

    work page doi:10.1109/cvpr.2016.90 2016

  38. [38]

    arXiv preprint arXiv:1508.01991 (2015)

    Zhiheng Huang, Wei Xu, & Kai Yu. Bidirectional LSTM-CRF Models for Sequence Tagging, (2015), doi: 10.48550/arXiv.1508.01991

    work page doi:10.48550/arxiv.1508.01991 2015

  39. [39]

    Zhang Ban, Xiao-Bo Li, Xun Yang, et al., Mock Observations for the CSST Mission: End-to-end Performance Modeling of Optical System, Res. Astron. Astrophys., 26 (2) (2026), Article 024002, doi: 10.1088/1674-4527/ae20fb

    work page doi:10.1088/1674-4527/ae20fb 2026

  40. [40]

    2010, Monthly Notices of the Royal Astronomical Society, 408, 1181, doi: 10.1111/j.1365-2966.2010.17197.x

    M. Henrion, D. J. Mortlock, D. J. Hand, A. Gandy, A Bayesian approach to star–galaxy classification, Mon. Not. R. Astron. Soc., 412 (4) (2011), pp. 2286-2302, doi: 10.1111/j.1365-2966.2010.18055.x

    work page doi:10.1111/j.1365-2966.2010.18055.x 2011

  41. [41]

    Bertin, Automated Morphometry with SExtractor and PSFEx, ADASS XX, 442 (2011), p

    E. Bertin, Automated Morphometry with SExtractor and PSFEx, ADASS XX, 442 (2011), p. 435

    2011

  42. [42]

    Richard C Tolman, Effect of Inhomogeneity on Cosmological Models, Proc Natl Acad Sci U S A., 20 (3) (1934), pp. 169-176

    1934