pith. sign in

arxiv: 2605.19649 · v2 · pith:QE7NI5OFnew · submitted 2026-05-19 · 💻 cs.CV

CAD-Free Learning of Spacecraft Pose Estimators via NeRF-Based Augmentations

Antoine Legrand , Renaud Detry , Christophe De Vleeschouwer This is my paper

Pith reviewed 2026-05-21 07:37 UTC · model grok-4.3

classification 💻 cs.CV
keywords spacecraft pose estimationNeRF augmentationCAD-freedata augmentationcomputer visionneural radiance fieldon-orbit imagery
0
0 comments X

The pith

NeRF-based augmentations enable training accurate spacecraft pose estimators from only 25 to 400 real images without CAD models.

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

The paper seeks to reduce reliance on large synthetic CAD datasets for training spacecraft pose estimators. Such datasets are costly to create and fail to capture real lighting and material properties, limiting applicability to known targets and causing poor real-world performance. The proposed method learns a Neural Radiance Field from a small number of real images and uses it to produce a much larger set of augmented images with consistent geometry but varied appearances and views. Experiments demonstrate that this suffices to train accurate estimators even with severe illumination changes. A successful outcome would expand pose estimation to uncooperative spacecraft lacking detailed 3D models.

Core claim

By training a Neural Radiance Field on 25 to 400 real images of a target spacecraft and generating geometrically consistent viewpoint and appearance augmentations, accurate target-specific pose estimators can be learned without a CAD model or large synthetic datasets, and this also improves generalization when applied to CAD data.

What carries the argument

The NeRF-based image augmentation that learns a Neural Radiance Field of the spacecraft to generate diverse yet geometrically accurate training images for pose estimation.

Load-bearing premise

A Neural Radiance Field trained on a few hundred real images can produce augmentations free of systematic pose label biases that still generalize to unseen real on-orbit images.

What would settle it

If a pose estimator trained solely on the NeRF-augmented data shows high error rates when tested on real images captured separately under novel lighting conditions, the claim would be falsified.

Figures

Figures reproduced from arXiv: 2605.19649 by Antoine Legrand, Christophe De Vleeschouwer, Renaud Detry.

Figure 1
Figure 1. Figure 1: Our NeRF-based offline dataset augmentation method. Instead of training a pose estimation network on dataset S Orig limited in terms of both volume and diversity, we train the network on an augmented set S Augm that combines S Orig with a set S NeRF. This set contains images generated through a NeRF trained on S Orig under novel pose labels and randomized appearance conditions. This dataset augmentation en… view at source ↗
Figure 2
Figure 2. Figure 2: Image generation through a Neural Radiance Field (see section [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Method Overview. We first learn two neural radiance fields [PITH_FULL_IMAGE:figures/full_fig_p004_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: Architectural overview of SPNv2 [16]. The network extracts multi-scale feature maps using an EfficientNet back￾bone and a BiFPN. They are processed through three heads: (i) a pose-regression head, (ii) a keypoint head whose predictions are used by a PnP solver to compute the pose, and (iii) a segmen￾tation head providing auxiliary supervision. The model outputs a hybrid pose combining the regressed transla… view at source ↗
read the original abstract

Spacecraft pose estimation networks require tens of thousands of CAD-rendered images to be trained. This reliance on synthetic CAD data (i) limits applicability to targets with reliable geometry prior, excluding uncooperative or poorly documented spacecraft, and (ii) causes poor generalization to real on-orbit conditions due to unrealistic illumination and material appearance. This paper introduces a NeRF-based image augmentation method that enables the learning of spacecraft pose estimators from only a few tens to a few hundreds of images. The method learns a Neural Radiance Field of the target and generates a large, diverse dataset through geometrically-consistent viewpoint and appearance augmentation. This augmented dataset enables the training of accurate target-specific pose estimators without requiring a CAD model or large synthetic datasets. Experiments show that our approach supports the training of accurate pose estimators from only 25 to 400 realistic images, even under severe illumination variations. When applied on large CAD-based synthetic datasets, the NeRF-based augmentation also enhances out-of-domain generalization, yielding improved robustness to real on-orbit conditions.

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

2 major / 1 minor

Summary. The manuscript presents a NeRF-based augmentation pipeline that learns a Neural Radiance Field from 25–400 real images of a spacecraft target and uses the resulting model to synthesize large numbers of geometrically consistent, appearance-augmented training images. These images are then used to train a target-specific pose estimator without any CAD model or large synthetic dataset, with the claim that the resulting estimators remain accurate even under severe illumination variations and generalize better to real on-orbit imagery.

Significance. If the empirical results hold, the work would meaningfully expand the applicability of learned pose estimation to uncooperative or poorly documented spacecraft for which CAD models are unavailable. The use of NeRF to produce pose-labeled augmentations directly from real images is a practical way to reduce the sim-to-real gap and could be adopted in operational space-vision pipelines.

major comments (2)
  1. [Abstract] Abstract: the central claim that accurate pose estimators can be trained from only 25–400 realistic images rests on the assumption that the input images already possess sufficiently accurate camera poses for NeRF optimization. The abstract gives no indication of how these poses are obtained or validated (e.g., via COLMAP-style SfM) under the stated severe illumination variations, where feature matching is expected to degrade. Any systematic error in the initial poses would propagate into the learned geometry and into every rendered augmentation’s pose label, directly undermining the downstream estimator.
  2. [Abstract] Abstract / Experiments (implied): the abstract asserts “positive experimental outcomes” and “accurate pose estimators” yet supplies no quantitative metrics, error bars, dataset sizes, or ablation results. Without these numbers it is impossible to judge whether the reported performance is load-bearing for the 25–400-image regime or merely anecdotal.
minor comments (1)
  1. [Abstract] The abstract would be strengthened by the inclusion of at least one concrete performance figure (e.g., median rotation/translation error on a held-out real test set) so readers can immediately gauge the magnitude of the improvement.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the constructive and insightful comments. We address each major comment point by point below and have revised the manuscript to improve clarity where the feedback identifies gaps in presentation.

read point-by-point responses

  1. Referee: [Abstract] Abstract: the central claim that accurate pose estimators can be trained from only 25–400 realistic images rests on the assumption that the input images already possess sufficiently accurate camera poses for NeRF optimization. The abstract gives no indication of how these poses are obtained or validated (e.g., via COLMAP-style SfM) under the stated severe illumination variations, where feature matching is expected to degrade. Any systematic error in the initial poses would propagate into the learned geometry and into every rendered augmentation’s pose label, directly undermining the downstream estimator.

    Authors: We agree that the abstract should make the pose acquisition step explicit to avoid ambiguity. The full manuscript (Section 3.1) explains that initial camera poses are recovered via COLMAP structure-from-motion applied directly to the small set of real images before NeRF optimization begins. We have revised the abstract to include a concise statement that poses are obtained through SfM on the input images. On the question of illumination-induced degradation, our experiments (Section 4.3) demonstrate that the subsequent NeRF training and appearance augmentation remain effective even when initial SfM poses contain moderate noise; the geometric consistency enforced by the radiance field mitigates propagation of small errors into the final pose labels. We have added a short paragraph in the revised manuscript discussing this robustness. revision: yes

  2. Referee: [Abstract] Abstract / Experiments (implied): the abstract asserts “positive experimental outcomes” and “accurate pose estimators” yet supplies no quantitative metrics, error bars, dataset sizes, or ablation results. Without these numbers it is impossible to judge whether the reported performance is load-bearing for the 25–400-image regime or merely anecdotal.

    Authors: We acknowledge that the original abstract is high-level and does not embed specific numbers. To address this, we have updated the abstract to reference the key quantitative findings from our experiments (e.g., pose estimation accuracy achieved with 25–400 images and the corresponding improvements in generalization). The revised abstract now points readers to the detailed metrics, error bars, dataset sizes, and ablation studies that appear in Section 4, thereby making the performance claims more concrete while respecting abstract length limits. revision: yes

Circularity Check

0 steps flagged

NeRF augmentation pipeline is empirically self-contained with no load-bearing circular steps

full rationale

The paper describes an empirical pipeline: capture a small set of real images, train a NeRF, render geometrically consistent augmentations with known poses, and train a pose estimator. No equations, fitted parameters renamed as predictions, or self-citation chains are presented in the provided text that reduce the central claim to its inputs by construction. The method relies on the independent assumption that NeRF can produce useful augmentations from limited real views, which is externally testable and not tautological. Minor self-citation risk is possible in related work but is not load-bearing here.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 0 invented entities

The approach rests on the assumption that NeRF can faithfully reconstruct spacecraft geometry and appearance from a small number of real images and that the generated augmentations preserve accurate pose labels.

axioms (1)
  • domain assumption A Neural Radiance Field trained on a few tens to hundreds of images of a spacecraft produces views whose 6-DoF pose labels remain accurate for downstream supervised training.
    Invoked when the paper states that the augmented dataset enables accurate pose estimators.

pith-pipeline@v0.9.0 · 5712 in / 1281 out tokens · 24938 ms · 2026-05-21T07:37:01.990573+00:00 · methodology

Review history (2 revisions) →

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

55 extracted references · 55 canonical work pages · 1 internal anchor

  1. [1]

    The darpa phoenix spacecraft servic- ing program: Overview and plans for risk reduction

    Carl Glen Henshaw. The darpa phoenix spacecraft servic- ing program: Overview and plans for risk reduction. In i-SAIRAS. European Space Agency, 2014

    work page 2014

  2. [2]

    Performance of northrop grumman’s mission extension vehicle (mev) rpo imagers at geo

    Matt Pyrak and Joseph Anderson. Performance of northrop grumman’s mission extension vehicle (mev) rpo imagers at geo. InAutonomous Systems: Sensors, Processing and Se- curity for Ground, Air , Sea and Space V ehicles and Infras- tructure 2022, volume 12115, pages 64–82. SPIE, 2022. doi: 10.1117/12.2631524

    work page doi:10.1117/12.2631524 2022

  3. [3]

    Forshaw, Guglielmo S

    Jason L. Forshaw, Guglielmo S. Aglietti, Nimal Navarathi- nam, Haval Kadhem, Thierry Salmon, Aurélien Pisseloup, Eric Joffre, Thomas Chabot, Ingo Retat, Robert Axthelm, Simon Barraclough, Andrew Ratcliffe, Cesar Bernal, François Chaumette, Alexandre Pollini, and Willem H. Steyn. Removedebris: An in-orbit active debris removal demonstration mission.Acta As...

    work page

  4. [4]

    doi: 10.1016/j.actaastro.2016.06

    ISSN 0094-5765. doi: 10.1016/j.actaastro.2016.06. 018

    work page doi:10.1016/j.actaastro.2016.06 2016

  5. [5]

    The clearspace-1 mission: Esa and clearspace team up to re- move debris

    Robin Biesbroek, Sarmad Aziz, Andrew Wolahan, Ste- fano Cipolla, Muriel Richard-Noca, and Luc Piguet. The clearspace-1 mission: Esa and clearspace team up to re- move debris. InProc. 8th Eur . Conf. Sp. Debris, pages 1–3, 2021

    work page 2021

  6. [6]

    PhD thesis, Technische Universität München, 2016

    Jacopo Ventura.Autonomous proximity operations for noncooperative space targets. PhD thesis, Technische Universität München, 2016

    work page 2016

  7. [7]

    A review of cooperative and uncooperative spacecraft pose determination techniques for close-proximity operations.Progress in Aerospace Sci- ences, 93:53–72, 2017

    Roberto Opromolla, Giancarmine Fasano, Giancarlo Rufino, and Michele Grassi. A review of cooperative and uncooperative spacecraft pose determination techniques for close-proximity operations.Progress in Aerospace Sci- ences, 93:53–72, 2017. doi: 10.1016/j.paerosci.2017.07. 001

    work page doi:10.1016/j.paerosci.2017.07 2017

  8. [8]

    Pose estimation for spacecraft relative navigation using model-based algorithms.IEEE Transactions on Aerospace and Electronic Systems, 53(1): 431–447, 2017

    Roberto Opromolla, Giancarmine Fasano, Giancarlo Rufino, and Michele Grassi. Pose estimation for spacecraft relative navigation using model-based algorithms.IEEE Transactions on Aerospace and Electronic Systems, 53(1): 431–447, 2017. doi: 10.1109/TAES.2017.2650785

    work page doi:10.1109/taes.2017.2650785 2017

  9. [9]

    Uncooperative spacecraft pose estimation using an infrared camera during proximity operations

    Jian-Feng Shi, Steve Ulrich, Stephane Ruel, and Martin Anctil. Uncooperative spacecraft pose estimation using an infrared camera during proximity operations. InAIAA SPACE 2015 conference and exposition, page 4429, 2015. doi: 10.2514/6.2015-4429

    work page doi:10.2514/6.2015-4429 2015

  10. [10]

    Pose estimation and tracking of non-cooperative rocket bodies using time-of-flight cameras.Acta Astronau- tica, 139:165–175, 2017

    Harvey Gómez Martínez, Gabriele Giorgi, and Bernd Eiss- feller. Pose estimation and tracking of non-cooperative rocket bodies using time-of-flight cameras.Acta Astronau- tica, 139:165–175, 2017. doi: 10.1016/j.actaastro.2017.07. 002

    work page doi:10.1016/j.actaastro.2017.07 2017

  11. [11]

    Towards bridging the space domain gap for satellite pose estimation using event sensing

    Mohsi Jawaid, Ethan Elms, Yasir Latif, and Tat-Jun Chin. Towards bridging the space domain gap for satellite pose estimation using event sensing. In2023 IEEE Interna- tional Conference on Robotics and Automation (ICRA), pages 11866–11873, 2023. doi: 10.1109/ICRA48891. 2023.10160531

    work page doi:10.1109/icra48891 2023

  12. [12]

    Vincenzo Pesce, Michèle Lavagna, and Riccardo Bevilac- qua. Stereovision-based pose and inertia estimation of unknown and uncooperative space objects.Advances in Preprint– CAD-FreeLearning ofSpacecraftPoseEstimators viaNeRF-BasedAugmentations10 Space Research, 59(1):236–251, 2017. doi: 10.1016/j.asr. 2016.10.002

    work page doi:10.1016/j.asr 2017

  13. [13]

    Pose estimation for non-cooperative spacecraft rendezvous using convolutional neural networks

    Sumant Sharma, Connor Beierle, and Simone D’Amico. Pose estimation for non-cooperative spacecraft rendezvous using convolutional neural networks. In2018 IEEE Aerospace Conference, pages 1–12. IEEE, 2018

    work page 2018

  14. [14]

    Satellite pose estimation challenge: Dataset, competition design, and results.IEEE Transactions on Aerospace and Electronic Systems, 56(5):4083–4098, 2020

    Mate Kisantal, Sumant Sharma, Tae Ha Park, Dario Izzo, Marcus Märtens, and Simone D’Amico. Satellite pose estimation challenge: Dataset, competition design, and results.IEEE Transactions on Aerospace and Electronic Systems, 56(5):4083–4098, 2020

    work page 2020

  15. [15]

    Speed+: Next-generation dataset for spacecraft pose estimation across domain gap

    Tae Ha Park, Marcus Märtens, Gurvan Lecuyer, Dario Izzo, and Simone D’Amico. Speed+: Next-generation dataset for spacecraft pose estimation across domain gap. In2022 IEEE Aerospace Conference (AERO), pages 1–15. IEEE, 2022

    work page 2022

  16. [16]

    Satellite pose estimation with deep landmark regression and nonlinear pose refinement

    Bo Chen, Jiewei Cao, Alvaro Parra, and Tat-Jun Chin. Satellite pose estimation with deep landmark regression and nonlinear pose refinement. InProceedings of the IEEE/CVF international conference on computer vision workshops, pages 0–0, 2019

    work page 2019

  17. [17]

    Robust multi-task learning and online refinement for spacecraft pose estima- tion across domain gap.Advances in Space Research, 73 (11):5726–5740, 2024

    Tae Ha Park and Simone D’Amico. Robust multi-task learning and online refinement for spacecraft pose estima- tion across domain gap.Advances in Space Research, 73 (11):5726–5740, 2024

    work page 2024

  18. [18]

    Domain generalization for in-orbit 6d pose estimation.Journal of Aerospace Information Systems, 22(11):938–947, 2025

    Antoine Legrand, Renaud Detry, and Christophe De Vleeschouwer. Domain generalization for in-orbit 6d pose estimation.Journal of Aerospace Information Systems, 22(11):938–947, 2025

    work page 2025

  19. [19]

    Robotic testbed for rendezvous and optical navigation: Multi-source calibration and machine learning use cases,

    Tae Ha Park, Juergen Bosse, and Simone D’Amico. Robotic testbed for rendezvous and optical navigation: Multi-source calibration and machine learning use cases. arXiv preprint arXiv:2108.05529, 2021

    work page arXiv 2021

  20. [20]

    Space debris generation in geo: Space materials testing and eval- uation.Acta Astronautica, 192:258–275, 2022

    Sophie Duzellier, Paulo Gordo, Rui Melicio, Duarte Valério, Mark Millinger, and António Amorim. Space debris generation in geo: Space materials testing and eval- uation.Acta Astronautica, 192:258–275, 2022

    work page 2022

  21. [21]

    Nerf: Representing scenes as neural radiance fields for view syn- thesis.Communications of the ACM, 65(1):99–106, 2021

    Ben Mildenhall, Pratul P Srinivasan, Matthew Tancik, Jonathan T Barron, Ravi Ramamoorthi, and Ren Ng. Nerf: Representing scenes as neural radiance fields for view syn- thesis.Communications of the ACM, 65(1):99–106, 2021

    work page 2021

  22. [22]

    Antoine Legrand, Renaud Detry, and Christophe De Vleeschouwer. Leveraging neural radiance fields for pose estimation of an unknown space object dur- ing proximity operations.IEEE International Confer- ence on Space Robotics (ISpaRo) Workshop, available at arXiv:2405.12728, 2024

    work page arXiv 2024

  23. [23]

    Domain generalization for 6D pose estimation through NeRF-based image synthesis,

    Antoine Legrand, Renaud Detry, and Christophe De Vleeschouwer. Domain generalization for 6d pose esti- mation through nerf-based image synthesis.arXiv preprint arXiv:2407.10762, 2024

    work page arXiv 2024

  24. [24]

    Satellite pose estimation competition 2021: Results and analyses.Acta Astronautica, 204:640–665, 2023

    Tae Ha Park, Marcus Märtens, Mohsi Jawaid, Zi Wang, Bo Chen, Tat-Jun Chin, Dario Izzo, and Simone D’Amico. Satellite pose estimation competition 2021: Results and analyses.Acta Astronautica, 204:640–665, 2023

    work page 2021

  25. [25]

    Perspective transformation data augmentation for object detection.IEEE Access, 8:4935–4943, 2019

    Ke Wang, Bin Fang, Jiye Qian, Su Yang, Xin Zhou, and Jie Zhou. Perspective transformation data augmentation for object detection.IEEE Access, 8:4935–4943, 2019

    work page 2019

  26. [26]

    Two-level data augmentation for calibrated multi- view detection

    Martin Engilberge, Haixin Shi, Zhiye Wang, and Pascal Fua. Two-level data augmentation for calibrated multi- view detection. InProceedings of the IEEE/CVF Winter Conference on Applications of Computer Vision, pages 128–136, 2023

    work page 2023

  27. [27]

    Generative spatiotemporal data aug- mentation.arXiv preprint arXiv:2512.12508, 2025

    Jinfan Zhou, Lixin Luo, Sungmin Eum, Heesung Kwon, and Jeong Joon Park. Generative spatiotemporal data aug- mentation.arXiv preprint arXiv:2512.12508, 2025

    work page arXiv 2025

  28. [28]

    Poseaug: A differentiable pose augmentation framework for 3d hu- man pose estimation

    Kehong Gong, Jianfeng Zhang, and Jiashi Feng. Poseaug: A differentiable pose augmentation framework for 3d hu- man pose estimation. InProceedings of the IEEE/CVF conference on computer vision and pattern recognition, pages 8575–8584, 2021

    work page 2021

  29. [29]

    Nerfmentation: Im- proving monocular depth estimation with nerf-based data augmentation

    Casimir Feldmann, Niall Siegenheim, Nikolas Hars, Lovro Rabuzin, Mert Ertugrul, Luca Wolfart, Marc Pollefeys, Zuria Bauer, and Martin R Oswald. Nerfmentation: Im- proving monocular depth estimation with nerf-based data augmentation. InEuropean Conference on Computer Vi- sion, pages 92–108. Springer, 2024

    work page 2024

  30. [30]

    Towards viewpoint- invariant visual recognition via adversarial training

    Shouwei Ruan, Yinpeng Dong, Hang Su, Jianteng Peng, Ning Chen, and Xingxing Wei. Towards viewpoint- invariant visual recognition via adversarial training. In Proceedings of the IEEE/CVF International Conference on Computer Vision, pages 4709–4719, 2023

    work page 2023

  31. [31]

    Neural-sim: Learning to generate training data with nerf

    Yunhao Ge, Harkirat Behl, Jiashu Xu, Suriya Gunasekar, Neel Joshi, Yale Song, Xin Wang, Laurent Itti, and Vibhav Vineet. Neural-sim: Learning to generate training data with nerf. InEuropean Conference on Computer Vision, pages 477–493. Springer, 2022

    work page 2022

  32. [32]

    NeRF-based Spacecraft Reconstruction from Monocular Imagery Under Illumination Variability and Pose Uncertainty

    Antoine Legrand, Renaud Detry, and Christophe De Vleeschouwer. Nerf-based spacecraft reconstruc- tion from close-range monocular imagery under illumi- nation variability and pose uncertainty.arXiv preprint arXiv:2605.18447, 2026

    work page internal anchor Pith review Pith/arXiv arXiv 2026

  33. [33]

    Domain randomization for transferring deep neural networks from simulation to the real world

    Josh Tobin, Rachel Fong, Alex Ray, Jonas Schneider, Woj- ciech Zaremba, and Pieter Abbeel. Domain randomization for transferring deep neural networks from simulation to the real world. In2017 IEEE/RSJ international conference on intelligent robots and systems (IROS), pages 23–30. IEEE, 2017

    work page 2017

  34. [34]

    Training deep networks with synthetic data: Bridging the reality gap by domain randomization

    Jonathan Tremblay, Aayush Prakash, David Acuna, Mark Brophy, Varun Jampani, Cem Anil, Thang To, Eric Cam- eracci, Shaad Boochoon, and Stan Birchfield. Training deep networks with synthetic data: Bridging the reality gap by domain randomization. InProceedings of the IEEE conference on computer vision and pattern recognition workshops, pages 969–977, 2018

    work page 2018

  35. [35]

    An annotation saved is an annotation earned: Using fully synthetic training for object detection

    Stefan Hinterstoisser, Olivier Pauly, Hauke Heibel, Marek Martina, and Martin Bokeloh. An annotation saved is an annotation earned: Using fully synthetic training for object detection. InProceedings of the IEEE/CVF inter- national conference on computer vision workshops, pages 0–0, 2019

    work page 2019

  36. [36]

    Photo-realistic neural domain Preprint– CAD-FreeLearning ofSpacecraftPoseEstimators viaNeRF-BasedAugmentations11 randomization

    Sergey Zakharov, Rares, Ambrus, , Vitor Guizilini, Wadim Kehl, and Adrien Gaidon. Photo-realistic neural domain Preprint– CAD-FreeLearning ofSpacecraftPoseEstimators viaNeRF-BasedAugmentations11 randomization. InEuropean Conference on Computer Vision, pages 310–327. Springer, 2022

    work page 2022

  37. [37]

    Style augmentation: data augmentation via style randomization

    Philip TG Jackson, Amir Atapour Abarghouei, Stephen Bonner, Toby P Breckon, and Boguslaw Obara. Style augmentation: data augmentation via style randomization. InCVPR workshops, volume 6, pages 10–11, 2019

    work page 2019

  38. [38]

    Domain randomization and pyramid consistency: Simulation-to-real generalization without accessing target domain data

    Xiangyu Yue, Yang Zhang, Sicheng Zhao, Alberto Sangiovanni-Vincentelli, Kurt Keutzer, and Boqing Gong. Domain randomization and pyramid consistency: Simulation-to-real generalization without accessing target domain data. InProceedings of the IEEE/CVF interna- tional conference on computer vision, pages 2100–2110, 2019

    work page 2019

  39. [39]

    Robust and generalizable visual represen- tation learning via random convolutions.arXiv preprint arXiv:2007.13003, 2020

    Zhenlin Xu, Deyi Liu, Junlin Yang, Colin Raffel, and Marc Niethammer. Robust and generalizable visual represen- tation learning via random convolutions.arXiv preprint arXiv:2007.13003, 2020

    work page arXiv 2007

  40. [40]

    Network randomization: A simple technique for general- ization in deep reinforcement learning

    Kimin Lee, Kibok Lee, Jinwoo Shin, and Honglak Lee. Network randomization: A simple technique for general- ization in deep reinforcement learning. InInternational Conference on Learning Representations, 2020

    work page 2020

  41. [41]

    A fourier-based framework for domain gen- eralization

    Qinwei Xu, Ruipeng Zhang, Ya Zhang, Yanfeng Wang, and Qi Tian. A fourier-based framework for domain gen- eralization. InProceedings of the IEEE/CVF conference on computer vision and pattern recognition, pages 14383– 14392, 2021

    work page 2021

  42. [42]

    Bridging the domain gap in satellite pose estimation: a self-training approach based on geometrical constraints

    Zi Wang, Minglin Chen, Yulan Guo, Zhang Li, and Qifeng Yu. Bridging the domain gap in satellite pose estimation: a self-training approach based on geometrical constraints. IEEE Transactions on Aerospace and Electronic Systems, 2023

    work page 2023

  43. [43]

    Spacecraft pose estimation based on unsupervised domain adaptation and on a 3d-guided loss combination

    Juan Ignacio Bravo Pérez-Villar, Álvaro García-Martín, and Jesús Bescós. Spacecraft pose estimation based on unsupervised domain adaptation and on a 3d-guided loss combination. InEuropean Conference on Computer Vision, pages 37–52. Springer, 2022

    work page 2022

  44. [44]

    Utilization of fpga for onboard inference of landmark localization in cnn-based spacecraft pose estimation.Aerospace, 7(11):159, 2020

    Kiruki Cosmas and Asami Kenichi. Utilization of fpga for onboard inference of landmark localization in cnn-based spacecraft pose estimation.Aerospace, 7(11):159, 2020. doi: 10.3390/aerospace7110159

    work page doi:10.3390/aerospace7110159 2020

  45. [45]

    Towards employing fpga and asip acceleration to enable onboard ai/ml in space applica- tions

    Vasileios Leon, George Lentaris, Dimitrios Soudris, Simon Vellas, and Mathieu Bernou. Towards employing fpga and asip acceleration to enable onboard ai/ml in space applica- tions. In2022 IFIP/IEEE 30th International Conference on V ery Large Scale Integration (VLSI-SoC), pages 1–4. IEEE,

    work page

  46. [46]

    doi: 10.1109/VLSI-SoC54400.2022.9939566

    work page doi:10.1109/vlsi-soc54400.2022.9939566 2022

  47. [47]

    Nerf in the wild: Neural radiance fields for unconstrained photo collections

    Ricardo Martin-Brualla, Noha Radwan, Mehdi SM Saj- jadi, Jonathan T Barron, Alexey Dosovitskiy, and Daniel Duckworth. Nerf in the wild: Neural radiance fields for unconstrained photo collections. InProceedings of the IEEE/CVF conference on computer vision and pattern recognition, pages 7210–7219, 2021

    work page 2021

  48. [48]

    K- planes: Explicit radiance fields in space, time, and appear- ance

    Sara Fridovich-Keil, Giacomo Meanti, Frederik Rahbæk Warburg, Benjamin Recht, and Angjoo Kanazawa. K- planes: Explicit radiance fields in space, time, and appear- ance. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 12479– 12488, 2023

    work page 2023

  49. [49]

    Structure-from-motion revisited

    Johannes L Schonberger and Jan-Michael Frahm. Structure-from-motion revisited. InProceedings of the IEEE conference on computer vision and pattern recogni- tion, pages 4104–4113, 2016

    work page 2016

  50. [50]

    Adaptive neural- network-based unscented kalman filter for robust pose tracking of noncooperative spacecraft.Journal of Guid- ance, Control, and Dynamics, 46(9):1671–1688, 2023

    Tae Ha Park and Simone D’Amico. Adaptive neural- network-based unscented kalman filter for robust pose tracking of noncooperative spacecraft.Journal of Guid- ance, Control, and Dynamics, 46(9):1671–1688, 2023

    work page 2023

  51. [51]

    Autonomous formation flying for the prisma mission.Jour- nal of Spacecraft and Rockets, 44(3):671–681, 2007

    Eberhard Gill, Simone D’Amico, and Oliver Montenbruck. Autonomous formation flying for the prisma mission.Jour- nal of Spacecraft and Rockets, 44(3):671–681, 2007

    work page 2007

  52. [52]

    Efficientnet: Rethinking model scaling for convolutional neural networks

    Mingxing Tan and Quoc Le. Efficientnet: Rethinking model scaling for convolutional neural networks. InIn- ternational conference on machine learning, pages 6105–

    work page

  53. [53]

    Efficient- det: Scalable and efficient object detection

    Mingxing Tan, Ruoming Pang, and Quoc V Le. Efficient- det: Scalable and efficient object detection. InProceedings of the IEEE/CVF conference on computer vision and pat- tern recognition, pages 10781–10790, 2020

    work page 2020

  54. [54]

    Efficientpose: An efficient, accurate and scalable end-to-end 6d multi object pose estimation approach.arXiv preprint arXiv:2011.04307, 2020

    Yannick Bukschat and Marcus Vetter. Efficientpose: An efficient, accurate and scalable end-to-end 6d multi object pose estimation approach.arXiv preprint arXiv:2011.04307, 2020

    work page arXiv 2011

  55. [55]

    Ep n p: An accurate o (n) solution to the p n p problem.In- ternational journal of computer vision, 81:155–166, 2009

    Vincent Lepetit, Francesc Moreno-Noguer, and Pascal Fua. Ep n p: An accurate o (n) solution to the p n p problem.In- ternational journal of computer vision, 81:155–166, 2009

    work page 2009