pith. machine review for the scientific record. sign in

arxiv: 2605.06913 · v1 · submitted 2026-05-07 · 🌌 astro-ph.EP · astro-ph.IM· cs.LG

Recognition: no theorem link

You Only Stack Once (YOSO): A Motion-Filtered, Deep-Learning Framework for Detecting Faint Moving Sources

Nitya Pandey , C\'esar Fuentes , Pedro Bernardinelli , Valeria Fr\'ias , Colin Orion Chandler , David E. Trilling , Matthew J. Holman , Steven Stetzler
show 10 more authors
Dallin Spencer Hsing Wen Lin Luis E. Salazar Manzano Darin Ragozzine Ryder Strauss Mario Juri\'c Andrew J. Connolly Hayden Smotherman Scott S. Sheppard Kevin Napier
Authors on Pith no claims yet

Pith reviewed 2026-05-11 01:03 UTC · model grok-4.3

classification 🌌 astro-ph.EP astro-ph.IMcs.LG
keywords moving sourcesSolar System objectsTNO detectionGaussian Motion Filterastronomical image processingshift-and-stack alternativefaint object detectionwide-field surveys
0
0 comments X

The pith

YOSO's Gaussian Motion Filter detects faint moving Solar System objects by enhancing trails at the pixel level rather than testing discrete velocities.

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

The paper presents YOSO, a pipeline that integrates a Gaussian Motion Filter to enhance the signal-to-noise ratio of faint, slow-moving Solar System objects in wide-field surveys. Unlike traditional shift-and-stack methods that require discrete velocity trials, this filter operates at the pixel level to amplify trails while suppressing noise and static backgrounds. When tested on a subset of DEEP observations, it recovered 45 of 73 known objects, identified 11 new trans-Neptunian objects, and discovered 216 objects in the near Solar System, all with an extremely low false positive rate. The method is proposed as scalable for large surveys like LSST and adaptable to other motion detection tasks.

Core claim

YOSO replaces multiple shift-and-stack operations with a single Gaussian Motion Filter that amplifies motion trails across a range of rates, resulting in the recovery of 45 out of 73 previously detected objects, the discovery of 11 new TNOs, and 216 near-Solar-System objects in DEEP data, while maintaining an extremely low false positive rate by requiring detections to show trails consistent with point sources at the correct rate.

What carries the argument

The Gaussian Motion Filter (GMoF), which enhances signal-to-noise for objects with varying apparent motion rates by amplifying trails and suppressing random noise and static features at the pixel level.

If this is right

  • YOSO can be deployed on large surveys like LSST for extracting faint motion-dependent signals.
  • The framework can be adapted for exoplanet imaging through Angular Differential Imaging.
  • It supports near-Earth object detection for missions like NEO Surveyor.
  • It offers a versatile approach for motion-based signal enhancement in data-intensive astronomy.

Where Pith is reading between the lines

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

  • If the filter generalizes to other datasets without retuning, it could significantly reduce computational demands in future surveys.
  • The low false positive rate could enable more automated detection pipelines with less need for manual review.
  • Similar techniques might be applied to detect moving sources in other imaging domains such as video surveillance or medical scans.

Load-bearing premise

The Gaussian Motion Filter parameters chosen for the DEEP data will generalize without retuning to other surveys or object populations, and the low false-positive claim holds on full unvetted datasets.

What would settle it

Applying the method to the full unvetted DEEP dataset or another survey and observing a high rate of false positives that fail to correspond to real objects upon follow-up observations.

Figures

Figures reproduced from arXiv: 2605.06913 by Andrew J. Connolly, C\'esar Fuentes, Colin Orion Chandler, Dallin Spencer, Darin Ragozzine, David E. Trilling, Hayden Smotherman, Hsing Wen Lin, Kevin Napier, Luis E. Salazar Manzano, Mario Juri\'c, Matthew J. Holman, Nitya Pandey, Pedro Bernardinelli, Ryder Strauss, Scott S. Sheppard, Steven Stetzler, Valeria Fr\'ias.

Figure 1
Figure 1. Figure 1: On-sky positions of DECam field pointings for quadrant B1. The footprint consists of 29 field nights, cov￾ering approximately 30 square degrees. Each field received ∼3.8 hours of integration time with individual exposures of 120 s. Field centers are defined by the R.A. and Dec. of the first exposure in each long stare. Of the 29 field nights, four have been analyzed by us to date, these are shown in dark b… view at source ↗
Figure 2
Figure 2. Figure 2: Upper panels show a synthetic moving object at three representative epochs (image indices 0, 50, and 99) as it traverses the image sequence. The colored boxes indicate the same detector pixel sampled at each epoch, whose values con￾tribute to the light curve shown below. The lower panel dis￾plays the resulting single-pixel light curve, with pixel counts plotted as a function of image index. Blue points rep… view at source ↗
Figure 3
Figure 3. Figure 3: Comparison of two image-combination statistics applied to a planted moving object of magnitude m = 25.06. The left panel shows the coadded image, with a blue box indicating the pixel used to extract the light curve. The central panel presents the single-pixel light curve before (blue points) and after (vermilion crosses) applying the Gaussian Motion Filter (GMoF; Section 3.1). The dashed green line marks t… view at source ↗
Figure 4
Figure 4. Figure 4: Detections by our best-trained model marked with blue boxes labeled “Moving Obj,” and the model’s de￾tection confidence. Poorly subtracted stars, image artifacts, and cosmic rays were successfully rejected. Panel (b) also contains two satellite trails, which the model correctly ig￾nored. bounding box localizes the trail, it does not uniquely determine the direction or rate of motion. The trail lies along a… view at source ↗
Figure 5
Figure 5. Figure 5: Panels (a) and (c) show the objects flagged by the machine learning model n2720 and n3472. Panels (b) and (d) show these confirmed detections after shifting and stacking to their right rate of motion, now renamed as B1j YOSO 04 (m∼24.3) and B1j YOSO 08 (m∼24.9), respectively. 5. PHOTOMETRY 5.1. Magnitude Determination from Stacked Images [PITH_FULL_IMAGE:figures/full_fig_p007_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: Comparison between implanted magnitudes for synthetic objects their measured photometry. Points are color-coded by their motion rate (pixels/hour). The blue solid line shows the best-fit linear calibration, and the dashed black line represents a one-to-one relationship. 5.2. Magnitude Calibration We calibrated the instrumental magnitudes onto the reference system by fitting a linear relation between measur… view at source ↗
Figure 7
Figure 7. Figure 7: Detection efficiency as a function of magnitude difference relative to the DEEP survey limit for each an￾alyzed field night. Colored curves show the recovery effi￾ciency obtained with our method, while the shaded regions indicate the associated uncertainties. The black vertical line at ∆m = 0 marks the nominal DEEP detection limit re￾ported by Napier et al. (2024). The horizontal dashed line denotes the 25… view at source ↗
Figure 8
Figure 8. Figure 8: Cumulative magnitude distribution of Trans￾Neptunian Objects in the analyzed DEEP survey fields 1. The blue step curve shows all TNOs reported by the DEEP survey (Smotherman et al. 2024), while the orange step curve shows the subset of those objects recovered by our pipeline. Downward-pointing arrows mark the calibrated magnitudes of newly discovered TNOs identified by our method that were not reported in … view at source ↗
Figure 9
Figure 9. Figure 9: (Left) A newly detected object (m ∼ 20.5), not previously reported in the DEEP catalog, is barely noticeable in the average image. Based on its apparent motion, it is consistent with a main-belt asteroid. (Middle) The motion-filtered image enhances its trail, and YOSO flags a segment of it as n2756. (Right) The shift-and-stack refinement converges to a precise apparent motion, of ( ˙α, ˙δ) ≈ (−26.00, −10.4… view at source ↗
read the original abstract

We present You Only Stack Once (YOSO), an automated pipeline designed to detect faint, slow-moving Solar System objects in wide-field astronomical surveys. The pipeline integrates a novel Gaussian Motion Filter (GMoF) that operates at the pixel level to enhance signal-to-noise for objects exhibiting a range of apparent rates of motion. Unlike conventional shift-and-stack methods, which rely on discrete velocity trials, GMoF amplifies trails while suppressing random noise and static background features. Applied to a subset of DEEP observations from the Dark Energy Camera, YOSO recovered 45 out of 73 previously detected objects, as well as 11 new TNOs. It also discovered 216 objects in the near Solar System. Although alternative shift-and-stack methods are sensitive to objects about 0.88 magnitudes fainter, YOSO's false positive rate is extremely low, since it detects only sources that exhibit a trail and are consistent with a point source when shifted at the right rate. We show how this method can be deployed on large surveys like LSST, and adapted for other domains that require motion-based signal enhancement, including exoplanet imaging through Angular Differential Imaging (ADI), and near-Earth object (NEO) detection for missions like NEO Surveyor. YOSO thus provides a versatile, scalable approach for extracting faint, motion-dependent signals in the era of data-intensive astronomy.

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

3 major / 2 minor

Summary. The manuscript presents You Only Stack Once (YOSO), a deep-learning pipeline that incorporates a novel Gaussian Motion Filter (GMoF) operating at the pixel level to enhance trails from faint, slow-moving Solar System objects while suppressing noise and static features. Unlike discrete-velocity shift-and-stack approaches, GMoF is claimed to be more efficient for a range of apparent motions. Applied to a subset of DEEP Dark Energy Camera observations, the method recovers 45 of 73 previously known objects, identifies 11 new TNOs, and detects 216 near-Solar-System objects; it is stated to be 0.88 magnitudes less sensitive than alternatives but to possess an extremely low false-positive rate because detections require both a visible trail under GMoF and point-source consistency after the optimal shift. The work positions YOSO as scalable for LSST and adaptable to ADI exoplanet imaging and NEO surveys.

Significance. If the low false-positive rate and recovery performance are confirmed on fully blind, unvetted data and the GMoF parameters generalize without retuning, the approach could provide a practical, lower-compute alternative to conventional shift-and-stack pipelines for LSST-scale moving-object searches. The pixel-level motion filtering is a concrete technical contribution that could be integrated into existing survey processing frameworks.

major comments (3)
  1. [Abstract] Abstract: The assertion that YOSO's false-positive rate is 'extremely low' rests on the requirement that sources exhibit both a trail under GMoF and point-source morphology after optimal shift, yet no quantitative false-positive measurement (e.g., rate per square degree or per exposure) is supplied for fields containing no known moving sources. All reported performance numbers derive from a pre-selected subset already known to contain real objects, so the observed near-zero FP rate does not constrain the rate that would appear on the full blind DEEP survey or on LSST-scale data.
  2. [Abstract] Abstract and results description: The recovery of 45/73 objects and the discovery of 11 new TNOs plus 216 near-Solar-System objects are presented without accompanying details on data splits, cross-validation procedure, error bars on the recovery fraction, or a direct side-by-side comparison table against a standard shift-and-stack baseline on the same fields.
  3. [Abstract] Abstract: The claim that GMoF parameters chosen for DEEP data will generalize to other surveys or object populations is stated without supporting tests; the manuscript provides no ablation or transfer experiment showing performance when the filter is applied unchanged to a different instrument or cadence.
minor comments (2)
  1. [Abstract] Abstract: The sentence stating that alternative methods are 'sensitive to objects about 0.88 magnitudes fainter' is ambiguous; it should be clarified whether this means YOSO reaches 0.88 mag shallower or deeper than the alternatives.
  2. [Abstract] Abstract: The adaptability claims for Angular Differential Imaging and NEO Surveyor are mentioned only in passing; a brief concrete example or reference to how the GMoF would be re-parameterized for those domains would improve clarity.

Simulated Author's Rebuttal

3 responses · 0 unresolved

We thank the referee for their constructive comments, which help clarify the scope and limitations of our evaluation. We address each major point below and have revised the manuscript to improve the presentation of results and qualify our claims where necessary.

read point-by-point responses

  1. Referee: [Abstract] Abstract: The assertion that YOSO's false-positive rate is 'extremely low' rests on the requirement that sources exhibit both a trail under GMoF and point-source morphology after optimal shift, yet no quantitative false-positive measurement (e.g., rate per square degree or per exposure) is supplied for fields containing no known moving sources. All reported performance numbers derive from a pre-selected subset already known to contain real objects, so the observed near-zero FP rate does not constrain the rate that would appear on the full blind DEEP survey or on LSST-scale data.

    Authors: We agree that the current evaluation on a pre-selected subset does not directly constrain FP rates on fully blind data. In the revised manuscript we have added a quantitative FP analysis using synthetic source injections into fields without known moving objects, reporting rates per square degree. This analysis supports the low-FP behavior arising from the dual trail-plus-point-source requirement while explicitly noting that full blind DEEP or LSST-scale rates remain to be measured on larger unvetted datasets. revision: yes

  2. Referee: [Abstract] Abstract and results description: The recovery of 45/73 objects and the discovery of 11 new TNOs plus 216 near-Solar-System objects are presented without accompanying details on data splits, cross-validation procedure, error bars on the recovery fraction, or a direct side-by-side comparison table against a standard shift-and-stack baseline on the same fields.

    Authors: The reported recovery figures are deterministic results on a fixed observational subset rather than outputs of a trained model, so no cross-validation splits were performed. We have expanded the methods section with the precise field selection criteria, the total area and number of exposures processed, and bootstrap-derived uncertainties on the recovery fraction. A new comparison table now shows YOSO versus a standard shift-and-stack pipeline run on identical fields, quantifying the sensitivity versus FP-rate trade-off. revision: yes

  3. Referee: [Abstract] Abstract: The claim that GMoF parameters chosen for DEEP data will generalize to other surveys or object populations is stated without supporting tests; the manuscript provides no ablation or transfer experiment showing performance when the filter is applied unchanged to a different instrument or cadence.

    Authors: GMoF parameters are chosen from the physical range of apparent motions and the Gaussian model of trail profiles rather than being fit to DEEP data alone. We have added an ablation study varying the Gaussian width on the DEEP fields and a short discussion of the physical motivation for parameter selection. To avoid overstatement we have removed the generalization claim from the abstract and now state that transfer performance to other instruments or cadences is expected but has not yet been tested. revision: partial

Circularity Check

0 steps flagged

No circularity: performance claims rest on external DEEP data application

full rationale

The paper introduces the YOSO pipeline and GMoF filter as a novel method, then applies it directly to a subset of DEEP observations from the Dark Energy Camera. Reported results (45/73 recoveries, 11 new TNOs, 216 near-Solar-System objects) are outcomes of this external application rather than any fitted parameter or prediction derived by construction from the same inputs. No equations, self-definitions, or self-citation chains reduce the central claims to tautologies. The low false-positive assertion follows from the method's design criteria (trail detection plus point-source consistency after shift), which is independent of the performance numbers themselves. This is a standard self-contained validation against external benchmarks.

Axiom & Free-Parameter Ledger

0 free parameters · 2 axioms · 1 invented entities

The central claim rests on the assumption that pixel-level Gaussian filtering can reliably separate moving point sources from noise and static backgrounds across varying rates without introducing systematic biases. No explicit free parameters or invented physical entities are described in the abstract; the filter itself is the novel algorithmic component.

axioms (2)
  • domain assumption Moving Solar System objects appear as linear trails or point sources after appropriate shifting in difference images.
    Invoked implicitly when stating that the filter amplifies trails while the final detection requires consistency with a point source.
  • domain assumption Random noise and static background features can be suppressed by a motion-tuned Gaussian kernel without removing real signals.
    Core premise of the GMoF design stated in the abstract.
invented entities (1)
  • Gaussian Motion Filter (GMoF) no independent evidence
    purpose: Pixel-level enhancement of motion trails for faint moving sources
    New algorithmic component introduced by the paper; no independent physical evidence required as it is a signal-processing construct.

pith-pipeline@v0.9.0 · 5642 in / 1631 out tokens · 30826 ms · 2026-05-11T01:03:52.680609+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

31 extracted references · 6 canonical work pages

  1. [1]

    Alard, C., & Lupton, R. H. 1998, The Astrophysical Journal, 503, 325 Astropy Collaboration, Robitaille, T. P., Tollerud, E. J., et al. 2013, A&A, 558, A33, doi: 10.1051/0004-6361/201322068 Astropy Collaboration, Price-Whelan, A. M., Sip˝ ocz, B. M., et al. 2018, AJ, 156, 123, doi: 10.3847/1538-3881/aabc4f Astropy Collaboration, Price-Whelan, A. M., Lim, P...

    work page doi:10.1051/0004-6361/201322068 1998

  2. [2]

    H., Smotherman, H., Langford, Z., et al

    Bernardinelli, P. H., Smotherman, H., Langford, Z., et al. 2024, The Astronomical Journal, 167, 134

    2024

  3. [3]

    M., Trilling, D., Allen, R., et al

    Bernstein, G. M., Trilling, D., Allen, R., et al. 2004, The Astronomical Journal, 128, 1364

    2004

  4. [4]

    1996, Astronomy and astrophysics supplement series, 117, 393

    Bertin, E., & Arnouts, S. 1996, Astronomy and astrophysics supplement series, 117, 393

    1996

  5. [5]

    2021, astropy/photutils: 1.3.0, https://doi.org/10.5281/zenodo.5796924

    Bradley, L., Sip˝ ocz, B., Robitaille, T., Tollerud, E., et al. 2021, astropy/photutils: 1.3.0, https://doi.org/10.5281/zenodo.5796924

    work page doi:10.5281/zenodo.5796924 2021

  6. [6]

    2020, Zenodo

    Bradley, L., Sip˝ ocz, B., Robitaille, T., et al. 2020, Zenodo

    2020

  7. [7]

    A., Mahabal, A., Ye, Q., et al

    Duev, D. A., Mahabal, A., Ye, Q., et al. 2019, Monthly Notices of the Royal Astronomical Society, 486, 4158

    2019

  8. [8]

    Edgeworth, K. E. 1943, Journal of the British Astronomical Association, vol. 53, p. 181-188 (1943)., 53, 181

    1943

  9. [9]

    , keywords =

    Flaugher, B., Diehl, H. T., Honscheid, K., et al. 2015, AJ, 150, 150, doi: 10.1088/0004-6256/150/5/150

    work page doi:10.1088/0004-6256/150/5/150 2015

  10. [10]

    Follette, K. B. 2023, Publications of the Astronomical Society of the Pacific, 135, 093001

    2023

  11. [11]

    Fraser, W. C. 2025, in Machine Learning for Small Bodies in the Solar System (Elsevier), 229–271 Google. 2024, Google Colaboratory, https://colab.research.google.com/

    2025

  12. [12]

    R., Millman, K

    Harris, C. R., Millman, K. J., van der Walt, S. J., et al. 2020, Nature, 585, 357, doi: 10.1038/s41586-020-2649-2

    work page doi:10.1038/s41586-020-2649-2 2020

  13. [13]

    Hunter, J. D. 2007, Computing in science & engineering, 9, 90 Iglesias Alvarez, S., Diez Alonso, E., Sanchez Rodriguez, M. L., et al. 2023, Axioms, 12, 348

    2007

  14. [14]

    2023, Ultralytics YOLOv8, https://github.com/ultralytics/ultralytics

    Jocher, G., Chaurasia, A., & Qiu, J. 2023, Ultralytics YOLOv8, https://github.com/ultralytics/ultralytics

    2023

  15. [15]

    Kuiper, G. P. 1951, Proceedings of the National Academy of Sciences, 37, 717

    1951

  16. [16]

    2025, arXiv preprint arXiv:2502.15300

    Li, G., Lu, Z., Wang, J., & Wang, Z. 2025, arXiv preprint arXiv:2502.15300

    work page arXiv 2025

  17. [17]

    2014, in European conference on computer vision, Springer, 740–755

    Lin, T.-Y., Maire, M., Belongie, S., et al. 2014, in European conference on computer vision, Springer, 740–755

    2014

  18. [18]

    R., Abell, P

    Mainzer, A., Masiero, J. R., Abell, P. A., et al. 2023, The Planetary Science Journal, 4, 224

    2023

  19. [19]

    2010, in Proceedings of the 9th Python in Science Conference, ed

    McKinney, W. 2010, in Proceedings of the 9th Python in Science Conference, ed. S. van der Walt & J. Millman, 51 – 56

    2010

  20. [20]

    2023, Astronomy & Astrophysics, 680, A74

    Mesarcik, M., Boonstra, A.-J., Iacobelli, M., et al. 2023, Astronomy & Astrophysics, 680, A74

    2023

  21. [21]

    J., Lin, H

    Napier, K. J., Lin, H. W., Gerdes, D. W., et al. 2024, The Planetary Science Journal, 5, 50 13 NOIRLab. 2022, DECam Community Pipeline (PL206), https://noirlab.edu/science/data-services/ data-reduction-software/csdc-mso-pipelines/pl206

    2024

  22. [22]

    2020, Astronomy & Astrophysics, 636, A75

    Ostdiek, B., Necib, L., Cohen, T., et al. 2020, Astronomy & Astrophysics, 636, A75

    2020

  23. [23]

    J., & Vanderburg, A

    Shallue, C. J., & Vanderburg, A. 2018, The Astronomical Journal, 155, 94

    2018

  24. [24]

    J., Kalmbach, J

    Smotherman, H., Connolly, A. J., Kalmbach, J. B., et al. 2021, The Astronomical Journal, 162, 245

    2021

  25. [25]

    H., Portillo, S

    Smotherman, H., Bernardinelli, P. H., Portillo, S. K., et al. 2024, The Astronomical Journal, 167, 136

    2024

  26. [26]

    E., Gerdes, D

    Trilling, D. E., Gerdes, D. W., Juri´ c, M., et al. 2024, The Astronomical Journal, 167, 132

    2024

  27. [27]

    A., Fuentes, C., Gerdes, D

    Trujillo, C. A., Fuentes, C., Gerdes, D. W., et al. 2024, The Astronomical Journal, 167, 133

    2024

  28. [28]

    1992, in American Astronomical Society Meeting Abstracts, Vol

    Tyson, J., Guhathakurta, P., Bernstein, G., & Hut, P. 1992, in American Astronomical Society Meeting Abstracts, Vol. 181, 06–10

    1992

  29. [29]

    2014, Astronomical Data Analysis Software and Systems XXIII, 485, 379

    Valdes, F., Gruendl, R., et al. 2014, Astronomical Data Analysis Software and Systems XXIII, 485, 379

    2014

  30. [30]

    E., et al

    Virtanen, P., Gommers, R., Oliphant, T. E., et al. 2020, Nature Methods, doi: https://doi.org/10.1038/s41592-019-0686-2

    work page doi:10.1038/s41592-019-0686-2 2020

  31. [31]

    J., Kalmbach, J

    Whidden, P. J., Kalmbach, J. B., Connolly, A. J., et al. 2019, The Astronomical Journal, 157, 119

    2019