pith. sign in

arxiv: 2605.19179 · v1 · pith:KU7YJH6Znew · submitted 2026-05-18 · 🌌 astro-ph.EP · astro-ph.IM· cs.LG

A Cloud-Based Tool for Meteorite Recovery Using Drones and Machine Learning

Seamus L. Anderson , Hadrien A. R. Devillepoix , Lewis Lakerink , Sawitchaya Tippaya , Dale P. Giancono , Martin C. Towner , Iona Clemente , Martin Cup\'ak
show 25 more authors
Ashley F. Rogers John H. Fairweather Mia Walker Daniel Burgin Michael A. Frazer Sophie E. Deam Veronika Pazderov\'a Eleanor K. Sansom Benjamin A. D. Hartig Hely C. Branco Thomas Stevenson Isabella Hatty Anna Zappatini Anthony Lagain Tom Lovelock Auriane Egal Lucy Forman David Belton Simon Windsor Shibli Saleheen Asher Leslie Gregory B. Poole Andrew Langendam Rachel S. Kirby Andrew G. Tomkins
This is my paper

Pith reviewed 2026-05-20 07:04 UTC · model grok-4.3

classification 🌌 astro-ph.EP astro-ph.IMcs.LG
keywords meteorite recoverydronesmachine learningcloud-based toolmeteorite fallsAustralia
0
0 comments X

The pith

A cloud-based tool combines drone flights and machine learning to help recover meteorites from instrumentally observed falls.

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

This paper introduces a cloud-based tool designed to aid in the recovery of meteorites from falls that have been tracked by instruments. The system relies on drones to survey the area and machine learning to scan the images for meteorite candidates. Improvements over earlier versions are highlighted, along with results from deployments in South and Western Australia. A sympathetic reader would care because timely recovery of these samples can provide fresh insights into the solar system's composition before alteration occurs. The work shows both where the technique succeeds and where it still falls short in practical use.

Core claim

The authors present a cloud-based tool that uses drones and machine learning to help recover instrumentally observed meteorite falls, showcasing improvements upon previous iterations and detailing successes and limitations when applied to observed meteorite falls in South and Western Australia.

What carries the argument

The cloud-based platform for processing drone imagery with machine learning to detect meteorites.

If this is right

  • The system has been tested on real meteorite falls in Australia with some successes.
  • Limitations arise under varying terrain and lighting conditions.
  • The tool is offered to the meteoritics community for use on future falls.
  • Enhancements have been made to earlier versions of the recovery system.

Where Pith is reading between the lines

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

  • Similar drone and machine learning methods might assist in searching for other small objects on planetary surfaces.
  • Wider adoption could lead to more meteorites being recovered and studied before they degrade.
  • Combining this with additional sensors on drones could improve detection accuracy in challenging environments.

Load-bearing premise

The machine learning models can reliably detect meteorites in real-world drone imagery without excessive false positives or missed detections under varying terrain and lighting conditions.

What would settle it

Observing the system miss a confirmed meteorite or produce too many incorrect alerts during a drone survey of a known fall site would challenge the central claim.

Figures

Figures reproduced from arXiv: 2605.19179 by Andrew G. Tomkins, Andrew Langendam, Anna Zappatini, Anthony Lagain, Asher Leslie, Ashley F. Rogers, Auriane Egal, Benjamin A. D. Hartig, Dale P. Giancono, Daniel Burgin, David Belton, Eleanor K. Sansom, Gregory B. Poole, Hadrien A. R. Devillepoix, Hely C. Branco, Iona Clemente, Isabella Hatty, John H. Fairweather, Lewis Lakerink, Lucy Forman, Martin C. Towner, Martin Cup\'ak, Mia Walker, Michael A. Frazer, Rachel S. Kirby, Sawitchaya Tippaya, Seamus L. Anderson, Shibli Saleheen, Simon Windsor, Sophie E. Deam, Thomas Stevenson, Tom Lovelock, Veronika Pazderov\'a.

Figure 1
Figure 1. Figure 1: Data processing overview: RGB imaging capture, data upload, data processing, and candidates human review process. Using a mobile satellite communications terminal, ~30 Mb s -1 can be continuously uploaded. On a typical day, our drone surveying system (a DJI M300 equipped with a Zenmuse [PITH_FULL_IMAGE:figures/full_fig_p006_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: A screenshot of the Stage 4 function in the webapp at the DN230523_02 fall area. The background is an orthomosaic generated from low-resolution drone data (100 mm pixel-1 ). High￾resolution survey images (2 mm pixel-1 ) that contain stage 4 meteorite candidates, georeferenced and overlaid as aid to the user. Here the operators have visited the 5 targets, rejecting 4 of them (red markers), with the last one… view at source ↗
Figure 3
Figure 3. Figure 3: Drone survey images of all our recovered meteorites. All images are the same scale, with each side equal to 30.5 cm. CONCLUSIONS AND FUTURE WORK In this work, we present a comprehensive solution for using drones and machine learning to recover meteorite falls. The web application is accessible to the meteorite research community at https://find.gfo.rocks. In the subsections below, we detail some of our ant… view at source ↗
read the original abstract

We present a cloud-based tool that uses drones and machine learning to help recover instrumentally observed meteorite falls. We showcase a collection of improvements made upon previous iterations of our system, as well as detail the successes and limitations of this technique when applied to observed meteorite falls in South and Western Australia. This tool is available to the meteoritics research community upon request at https://find.gfo.rocks.

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

1 major / 1 minor

Summary. The manuscript presents a cloud-based tool integrating drone surveys with machine learning for recovering meteorites from instrumentally observed falls. It describes iterative improvements to the system and evaluates its application to real events in South and Western Australia, outlining both successes and limitations. The tool is made available to the community upon request.

Significance. If the full case studies include concrete detection statistics, terrain-specific performance data, and explicit false-positive analysis, the work offers a practical systems contribution to meteorite recovery. The emphasis on real-world deployment and community access strengthens its utility for the meteoritics community, though the absence of quantitative benchmarks in the provided abstract limits immediate assessment of impact.

major comments (1)
  1. [Abstract] Abstract: the claim to 'detail the successes and limitations' is not supported by any reported detection rates, precision-recall values, false-positive rates under varying lighting/terrain, or dataset sizes. Without these metrics the central performance assertions cannot be evaluated.
minor comments (1)
  1. The link to the tool (https://find.gfo.rocks) should be accompanied by a brief description of access requirements or usage examples to aid readers.

Simulated Author's Rebuttal

1 responses · 0 unresolved

We thank the referee for the positive summary and recommendation for minor revision. We address the single major comment below and will update the manuscript accordingly.

read point-by-point responses

  1. Referee: [Abstract] Abstract: the claim to 'detail the successes and limitations' is not supported by any reported detection rates, precision-recall values, false-positive rates under varying lighting/terrain, or dataset sizes. Without these metrics the central performance assertions cannot be evaluated.

    Authors: We agree that the abstract's phrasing overstates the quantitative detail provided. The full manuscript presents case studies from South and Western Australia that describe specific successes (e.g., recoveries achieved) and limitations (e.g., terrain and lighting challenges) through narrative examples rather than formal performance metrics such as precision-recall or false-positive rates. To resolve the mismatch, we will revise the abstract to replace 'detail the successes and limitations' with a more accurate description such as 'illustrate the practical successes and limitations observed' and, space permitting, add one or two key qualitative outcomes from the deployments. This change will be incorporated in the revised version. revision: yes

Circularity Check

0 steps flagged

No significant circularity in applied tool description

full rationale

The manuscript is an applied systems paper describing a deployed cloud-based drone+ML tool for meteorite recovery, including incremental improvements over prior versions and concrete outcomes from real observed falls in Australia. No equations, derivations, fitted parameters, or mathematical predictions appear in the provided text or abstract. The central claims rest on empirical field results and performance details rather than any internal reduction to inputs or self-citation chains, rendering the work self-contained against external benchmarks with no load-bearing circular steps.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

Abstract-only review yields no explicit free parameters, axioms, or invented entities; any ML training parameters are implicit but not detailed or fitted in the provided text.

pith-pipeline@v0.9.0 · 5752 in / 1176 out tokens · 40094 ms · 2026-05-20T07:04:31.864344+00:00 · methodology

discussion (0)

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

Lean theorems connected to this paper

Citations machine-checked in the Pith Canon. Every link opens the source theorem in the public Lean library.

What do these tags mean?
matches
The paper's claim is directly supported by a theorem in the formal canon.
supports
The theorem supports part of the paper's argument, but the paper may add assumptions or extra steps.
extends
The paper goes beyond the formal theorem; the theorem is a base layer rather than the whole result.
uses
The paper appears to rely on the theorem as machinery.
contradicts
The paper's claim conflicts with a theorem or certificate in the canon.
unclear
Pith found a possible connection, but the passage is too broad, indirect, or ambiguous to say the theorem truly supports the claim.

Reference graph

Works this paper leans on

25 extracted references · 25 canonical work pages

  1. [1]

    In: 2019 2nd International Conference on Signal Processing and Information Security (ICSPIS), pp.1–

    Meteorite hunting using deep learning and UA Vs. In: 2019 2nd International Conference on Signal Processing and Information Security (ICSPIS), pp.1–

    work page 2019

  2. [2]

    Meteoritics & Planetary Science 55(11): 2461–2471

    Machine learning for semi-automated meteorite recovery. Meteoritics & Planetary Science 55(11): 2461–2471. https://doi.org/10.1111/maps.13593 Anderson, S. L., Towner, M. C., Fairweather, J., Bland, P. A., Devillepoix, H. A. R., Sansom, E. K., Cupák, M., Shober, P. M., and Benedix, G.K

    work page doi:10.1111/maps.13593

  3. [3]

    The Astrophysical Journal Letters 930(2): L25

    Successful recovery of an observed meteorite fall using drones and machine learning. The Astrophysical Journal Letters 930(2): L25. https://doi.org/10.3847/2041-8213/ac66d4 18 Anderson, S. L., Devillepoix, H. A. R., Tomkins, A.G., Brand, H. E. A., Tait, A., Winchester, L., Soderholm, J., Sansom, E. K., Towner, M. C., and Cupák, M

    work page doi:10.3847/2041-8213/ac66d4 2041

  4. [4]

    Meteoritics & Planetary Science 58(10): 1385–1398

    Saint- Pierre-le-Viger (L5-6) from asteroid 2023 CX1 recovered in Normandy, France—220 years after the historic fall of L’Aigle (L6 breccia) in the neighborhood. Meteoritics & Planetary Science 58(10): 1385–1398. Bland, P. A., Spurný, P., Bevan, A. W. R., Howard, K. T., Towner, M. C., Benedix, G. K., Greenwood, R. C., Shrbený, L., Franchi, I. A., Deacon, ...

    work page 2023

  5. [5]

    Science 198(4318): 727–731

    Antarctica: A deep-freeze storehouse for meteorites. Science 198(4318): 727–731. https://doi.org/10.1126/science.198.4318.727 Ceplecha, Z

    work page doi:10.1126/science.198.4318.727

  6. [6]

    Available at: https://keras.io [Accessed 7 May 2026]

    Keras [software]. Available at: https://keras.io [Accessed 7 May 2026]. Citron, R. I., Shah, A., Sinha, S., Watkins, C. and Jenniskens, P.,

    work page 2026

  7. [7]

    In: 48th Annual Lunar and Planetary Science Conference, abstract #1964, pp.2528

    Meteorite recovery using an autonomous drone and machine learning. In: 48th Annual Lunar and Planetary Science Conference, abstract #1964, pp.2528. 19 Citron, R. I., Jenniskens, P., Watkins, C., Sinha, S., Shah, A., Raissi, C., Devillepoix, H., Albers, J., and Zolensky, M

    work page 1964

  8. [8]

    Meteoritics & Planetary Science 56(6): 1073–1085

    Recovery of meteorites using an autonomous drone and machine learning. Meteoritics & Planetary Science 56(6): 1073–1085. https://doi.org/10.1111/maps.13663 Colas, F., Zanda, B., Bouley, S., Jeanne, S., Malgoyre, A., Birlan, M., Blanpain, C., Gattacceca, J., Jorda, L., Lecubin, J. and others,

    work page doi:10.1111/maps.13663

  9. [9]

    In: Proceedings of the International Meteor Conference, 30th IMC, Sibiu, Romania, 2011, pp.9–12

    The status of the NASA all sky fireball network. In: Proceedings of the International Meteor Conference, 30th IMC, Sibiu, Romania, 2011, pp.9–12. Devillepoix, H. A. R., Sansom, E. K., Bland, P. A., Towner, M. C., Cupák, M., Howie, R. M., Jansen-Sturgeon, T., Cox, M. A., Hartig, B. A. D., Benedix, G. K., and Paxman, J

    work page 2011

  10. [10]

    Nature Astronomy 9(11): 1624–1637

    Catastrophic disruption of asteroid 2023 CX1 and implications for planetary defence. Nature Astronomy 9(11): 1624–1637. https://doi.org/10.1038/s41550-025-02659-8 Fries, M., and Fries, J

    work page doi:10.1038/s41550-025-02659-8 2023

  11. [11]

    Meteoritics & Planetary Science 49(11): 1989–1996

    Detection and rapid recovery of the Sutter’s Mill meteorite fall as a model for future recoveries worldwide. Meteoritics & Planetary Science 49(11): 1989–1996. 20 Gardiol, D., Barghini, D., Buzzoni, A., Carbognani, A., Di Carlo, M., Di Martino, M., Knapic, C., Londero, E., Pratesi, G., Rasetti, S. et al

    work page 1989

  12. [12]

    https://doi.org/10.1111/j.1945-5100.2011.01229.x Gritsevich, M., Lyytinen, E., Moilanen, J., Kohout, T., Dmitriev, V ., Lupovka, V ., Midtskogen, V ., Kruglikov, N., Ischenko, A., Yakovlev, G. et al

    work page doi:10.1111/j.1945-5100.2011.01229.x 1945

  13. [13]

    GitHub repository

    Drone-Footprints [software]. GitHub repository. Available at: https://github.com/spifftek70/Drone- Footprints/tree/86f918064d5684e895c66bd397c968a235bdafa5 [Accessed 7 May 2026]. Hill, P. J. A., Tunney, L. D., and Herd, C. D. K

    work page 2026

  14. [14]

    Meteoritics & Planetary Science 58(3): 421–432

    Application of drone-captured thermal imagery in aiding in the recovery of meteorites within a snow-covered strewn field. Meteoritics & Planetary Science 58(3): 421–432. https://doi.org/10.1111/maps.13963 Hofmann, B. A., and Gnos, E

    work page doi:10.1111/maps.13963

  15. [15]

    Geological Society of London Special Publications, 550(1): 469–492

    The meteoritic record of Arabia. Geological Society of London Special Publications, 550(1): 469–492. https://doi.org/10.1144/SP550-2024-8 Howie, R. M., Paxman, J., Bland, P. A., Towner, M. C., Cupak, M., Sansom, E. K., and Devillepoix, H.A.R

    work page doi:10.1144/sp550-2024-8 2024

  16. [16]

    Available at: https://github.com/ultralytics/ultralytics [Accessed 7 May 2026]

    Ultralytics YOLOv8 [software], version 8.0.0. Available at: https://github.com/ultralytics/ultralytics [Accessed 7 May 2026]. 21 King, A. J., Daly, L., Rowe, J., Joy, K. H., Greenwood, R. C., Devillepoix, H. A. R., Suttle, M. D., Chan, Q. H. S., Russell, S. S., Bates, H. C. et al

    work page 2026

  17. [17]

    In: Proceedings of the International Meteor Conference, 24th IMC, Oostmalle, Belgium, 2005, pp.53–62

    Polish fireball network. In: Proceedings of the International Meteor Conference, 24th IMC, Oostmalle, Belgium, 2005, pp.53–62. Pons Recasens, G.,

    work page 2005

  18. [18]

    Available at: https://github.com/roboflow/rf-detr [Accessed 7 May 2026]

    RF-DETR [software]. Available at: https://github.com/roboflow/rf-detr [Accessed 7 May 2026]. Russell, S. S., King, A. J., Bates, H. C., Almeida, N. V ., Greenwood, R. C., Daly, L., Joy, K. H., Rowe, J., Salge, T., Smith, C. L., et al

    work page 2026

  19. [19]

    Meteoritics & Planetary Science 60(2): 308–323

    Systematic meteorite collection in the Catalina Dense Collection area (Chile): Description and statistics. Meteoritics & Planetary Science 60(2): 308–323. https://doi.org/10.1111/maps.14307 Sadaka, C., Gattacceca, J., Dumas, F., Braucher, R., ASTER Team, Leya, I., Tauseef, M., Blard, P.-H., Bekaert, D., Füri, E., Zimmermann, L., Lagain, A., Devillepoix, H...

    work page doi:10.1111/maps.14307

  20. [20]

    Meteoritics & Planetary Science

    Terrestrial ages of meteorites from the Atacama Desert (Chile) and insights into the past meteorite flux to Earth. Meteoritics & Planetary Science. https://doi.org/10.1111/maps.70125 Sansom, E. K., Bland, P. A., Towner, M. C., Devillepoix, H. A. R., Cupák, M., Howie, R. M., Jansen-Sturgeon, T., Cox, M. A., Hartig, B. A. D., Paxman, J. P, et al

    work page doi:10.1111/maps.70125

  21. [21]

    Astronomy & Astrophysics 686: A67

    Atmospheric entry and fragmentation of the small asteroid 2024 BX1: Bolide trajectory, orbit, dynamics, light curve, and spectrum. Astronomy & Astrophysics 686: A67. Spurný, P., Bland, P. A., Shrbený, L., Borovička, J., Ceplecha, Z., Singleton, A., Bevan, A. W. R., Vaughan, D., Towner, M. C., McClafferty, T. P., et al

    work page 2024

  22. [22]

    Drag-line sensor for drone assisted meteorite detection

    “Drag-line sensor for drone assisted meteorite detection”. Bachelor thesis. Macquarie University. https://doi.org/10.25949/19428392 Towner, M. C., Jansen-Sturgeon, T., Cupak, M., Sansom, E. K., Devillepoix, H. A. R., Bland, P. A., Howie, R. M., Paxman, J. P., Benedix, G. K., and Hartig, B. A. D

    work page doi:10.25949/19428392

  23. [23]

    M., Llorca, J., Castro-Tirado, A

    https://doi.org/10.3847/PSJ/ac3df5 Trigo-Rodríguez, J. M., Llorca, J., Castro-Tirado, A. J., Ortiz, J. L., Docobo, J. A., and Fabregat, J

    work page doi:10.3847/psj/ac3df5

  24. [24]

    In: MetSoc 2023 – 86th Annual Meeting of the Meteoritical Society

    Recovery and planned study of the Saint-Pierre- Le-Viger meteorite: An achievement of the FRIPON/Vigie-Ciel citizen science program. In: MetSoc 2023 – 86th Annual Meeting of the Meteoritical Society. Zappatini, A., Gnos, E., Hofmann, B. A., Eggenberger, U., Kruttasch, P. M., Gfeller, F., Tauseef, M., Leya, I., Devillepoix, H. A. R., Sansom, E. K., Cupák, ...

    work page 2023

  25. [25]

    Meteoritics & Planetary Science

    Al-Khadhaf: The first camera-observed (H5–6) meteorite fall from Oman. Meteoritics & Planetary Science. https://doi.org/10.1111/maps.70110 Zender, J., Rudawska, R., Koschny, D., Drolshagen, G., Netjes, G.-J., Bosch, M., Bijl, R., Crevecoeur, R., and Bettonvil, F

    work page doi:10.1111/maps.70110