Pith. sign in

REVIEW 4 major objections 6 minor 80 references

Flying Co-Stereo: Enabling Long-Range Aerial Dense Mapping via Collaborative Stereo Vision of Dynamic-Baseline

T0 review · 4 major / 6 minor · reviewed 2026-08-07 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

Pith's one-line read Two cooperating drones form a wide-baseline stereo pair whose sparse triangulated landmarks rescale monocular depth estimates, yielding dense metric maps out to 70 meters with 2.3-9.7% relative error.

desk verdict A real two-UAV dynamic-baseline stereo system with solid engineering and a plausible but not fully secured 70 m dense-mapping claim; worth peer review with requests for stronger evaluation. read the letter →

arxiv 2506.00546 v1 pith:5NWY5ZCM submitted 2025-05-31 cs.RO

Zhaoying Wang , Xingxing Zuo , Wei Dong This is my paper
classification cs.RO
keywords FlyingCo-StereocollaborativestereovisiondynamicbaselineUAVswarmmappingsparse-to-densedepthexponentialfittingmonocularvisual-inertial-ranging
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

The paper tries to establish that two lightweight camera-and-IMU drones flying a few meters apart can act as a single wide-baseline stereo camera and produce dense, metric, long-range 3D maps without lidar. It claims the missing ingredient is not better sensors but a pipeline that estimates the changing baseline between the aircraft, keeps a shared set of features in view in real time, and then uses a handful of triangulated distant points to rescale a neural monocular depth map. If correct, small drone swarms could sense obstacles and terrain 70 m ahead with relative depth errors of 2.3-9.7%, more than tripling the range and quadrupling the coverage of a fixed-baseline onboard stereo camera.

What carries the argument

The load-bearing object is the exponential depth-fitting model in Eq. (13): a curve $\hat z = a\exp(b(\tilde d - c)) + d$ mapping the monocular network's up-to-scale depth $\tilde d$ to metric depth $\hat z$, with four parameters fitted from only 60-200 sparse landmarks using Ceres. This single global curve is what converts near-field VIO landmarks and long-range collaboratively triangulated landmarks into a full 640x480 dense metric depth image. Around it, the system stacks the DS-VIRE sliding-window estimator, which fuses visual PnP from infrared markers with IMU and UWB range factors and uses a bidirectional view differential for yaw, and the GP-SS hybrid, which uses periodic SuperPoint-SuperGlue matching as guidance and LK optical flow as continuous prediction. The machinery's job is to supply the sparse metric anchor points at long range and the accurate relative baseline that makes their triangulation stable.

What would settle it

Use a held-out outdoor scene with lidar ground truth: fit the four exponential parameters using only landmarks in the 30-50 m band, then compute per-pixel relative depth error on pixels beyond 70 m or in featureless regions; if those errors exceed the reported 2.3-9.7% band, or if the residual grows systematically with depth, the global-exponential assumption is false.

Watch

Extended reading notes

Core claim

Flying Co-Stereo's central claim is that a dynamic cross-agent baseline of 2-5 m between two UAVs turns their forward cameras into a wide-baseline stereo rig whose long-range parallax supports metric triangulation of distant co-visible landmarks, and that these landmarks are enough to convert an up-to-scale monocular depth prediction from DepthAnythingV2 into a dense metric depth map via a four-parameter exponential fit. The reported result is dense 3D mapping to 70 m with 2.3-9.7% relative error, with the wide-baseline configuration beating a 95 mm fixed stereo camera by up to 350% in range and 450% in coverage. The argument is carried by three coupled modules: the dual-spectrum visual-inertial-ranging estimator DS-VIRE for the time-varying baseline, the guidance-prediction feature association GP-SS for persistent co-visible points across dynamic viewpoints, and the exponential sparse-to-dense fitting that supplies the metric scale.

Load-bearing premise

The whole dense map is produced by one four-parameter exponential curve fitted to only 60-200 sparse landmarks, so the accuracy claim holds only if the monocular depth network's up-to-scale depth is globally exponential across every pixel of the 640x480 image, including far regions with no landmarks.

Editorial extensions

If this is right

  • Map range jumps from about 20 m to 70 m while the payload stays two lightweight camera-IMU-UWB drones, with no lidar.
  • Coverage area of the dense map is 331-450% larger than the onboard RealSense D455 stereo in the three outdoor test scenes.
  • Exponential fitting beats linear and quadratic alternatives consistently, cutting unidirectional Chamfer error by an average of 67% versus linear and 55% versus quadratic.
  • Real-time operation on a Jetson Xavier NX is feasible: GP-SS sustains near-30 Hz feature association instead of 13 Hz for full SuperGlue, while DS-MVDT keeps tracking the neighbor drone in 96-99% of frames across challenging outdoor scenes.
  • A baseline of 2-4 m balances triangulation stability against baseline-estimation error, with yaw estimation reaching 0.214 degrees mean absolute error.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • Editorial inference: if the exponential map between up-to-scale and metric depth is a property of the monocular network rather than of these particular scenes, the same sparse-to-dense recipe transfers to any future depth predictor by refitting the four parameters per frame.
  • Editorial inference: the measured baseline-length trade-off implies an active baseline controller could widen the drone separation when the scene is far and narrow it when the scene is near, a line the paper itself lists as future work.
  • Editorial inference: a held-out per-pixel test against lidar on regions without co-visible or VIO landmarks would reveal whether the global exponential fit, rather than the landmarks themselves, is carrying the 50-70 m far-field accuracy.
  • Editorial inference: stringing pairs so each drone tracks its predecessor and guides its successor would extend the moving baseline into a swarm-scale network, with coverage growing linearly with chain length; the paper names this as a future architecture.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

4 major / 6 minor

Summary. The paper proposes Flying Co-Stereo, a collaborative stereo mapping system in which two UAVs carrying forward-facing cameras form a wide, dynamic baseline. The system estimates the time-varying baseline with a dual-spectrum visual-inertial-ranging estimator (DS-VIRE), associates co-visible features across agents with a guidance-prediction SuperPoint/SuperGlue plus optical-flow scheme (GP-SS), and produces dense metric depth by exponentially fitting sparse triangulated landmarks to the up-to-scale depth of DepthAnythingV2 (Eq. 13). Experiments report baseline estimation MAE of 0.013 m and yaw MAE of 0.214 deg indoors, robust IR-marker tracking outdoors, and dense mapping up to 70 m with relative errors of 2.3% to 9.7% computed via a unidirectional Chamfer distance against R3LIVE ground truth, plus comparisons against a fixed-baseline RealSense stereo and two MVS methods.

Significance. If the central claim holds, the system is a meaningful step toward lightweight, long-range dense mapping for UAV swarms without LiDAR. The paper's strengths include real dual-UAV flight experiments, external motion-capture and R3LIVE ground truth, a concrete sparse-to-dense pipeline that goes beyond the linear fitting used in AB-VINS, and a clear analysis of triangulation stability versus baseline length. However, the headline dense-mapping accuracy rests on an evaluation design that does not yet validate every pixel of the reconstructed depth image, and the experimental basis is limited to very few frames per scene. These issues are fixable and do not invalidate the system concept, but they prevent the reported error numbers from being accepted at face value.

major comments (4)
  1. [Sec. VII-F, Eq. (17)] The dense-mapping accuracy metric uCD(P→G) is unidirectional: for each predicted point it queries only the nearest ground-truth point, so missing predicted geometry or arbitrarily sparse coverage in a depth band does not increase the error. The abstract and conclusion claim 2.3%–9.7% relative error and 70 m dense mapping based on this metric. Please also report a bidirectional Chamfer distance or a completeness/coverage metric (e.g., fraction of ground-truth points within a tolerance), and report per-pixel depth errors on the dense depth images rather than only on resampled predicted clouds.
  2. [Sec. VI-B, Eq. (13) and Sec. VII-F, Table IX] The four-parameter exponential transform is fit from only 60–200 sparse landmarks, and the evaluation does not separate landmark-supervised pixels from held-out pixels. As written, the reported depth-segment errors could reflect accuracy only near the sparse landmarks or in depth regimes where the landmarks are dense, not across the full 640×480 image out to 70 m. Please add a held-out validation protocol: fit the transform on one subset of landmarks (e.g., by depth band or spatial region) and evaluate the dense depth on the remaining landmarks or, better, on per-pixel R3LIVE depth, reporting errors per depth segment and per scene.
  3. [Sec. VII, Table VII and Sec. VII-F] The real-world dense-mapping evaluation uses only 4 camera frames per UAV per scene and a single flight per scene, with no repeated trials or error bars. For a claim of up to 70 m dense mapping and 331–450% coverage improvement, the evaluation should include longer sequences or repeated flights so that the reported uCD values and coverage areas are not artifacts of a single, short snippet. Please report means and variances over multiple runs or, failing that, explicitly state the single-run nature as a limitation.
  4. [Sec. VII-F, Table IX and Sec. VI-B] The paper does not state how the predicted point cloud P in Eq. (17) is generated from the dense depth map: whether all depth-image pixels are resampled, whether a depth threshold or confidence mask is applied, and how many points fall into each of the 0–10 m, 10–30 m, 30–50 m, and 50–70 m segments. Without this information, the per-segment relative errors in Table IX cannot be interpreted as dense reconstruction accuracy; they may reflect only a small subset of the image. Please specify the point-cloud generation procedure and report per-segment point counts and completeness.
minor comments (6)
  1. [Sec. IV-B, Eq. (3)] The visual residual in Eq. (3) is written as a two-row vector with a zero second row, but the context suggests it should be a 3-D position residual; please clarify the notation or remove the spurious zero component.
  2. [Sec. VI-A] The condition-number threshold used to discard poorly conditioned triangulations is never given; please report the actual threshold or the range used in the experiments.
  3. [Sec. VII-D, Table VI] The GP-SS total run-time is reported as "7.4/82.4 ms", which is ambiguous; please clarify which number is the LK-flow-only cost and which is the periodic SuperPoint/SuperGlue cost.
  4. [Sec. VII-F, Table IX] The relative error is defined as the ratio of uCD to the average depth of the segment, but the average depth values are not reported; please state them explicitly so the reader can verify the percentages.
  5. [Sec. VII and References] No code or data availability statement is provided; releasing the datasets and implementation would substantially strengthen reproducibility, especially given the small number of evaluation frames.
  6. [Throughout] There are typographical and formatting issues, including "V oxblox" in the Fig. 22 caption, "moshed WiFi" in Sec. VII-B, and inconsistent capitalization of "UAV" across the text; these should be cleaned up.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the dense mapping result is evaluated against external R3LIVE ground truth, and the exponential fit (Eq. 13) is a per-frame scaling model whose inputs are sparse landmarks, not the reported dense-map metric.

full rationale

The paper's derivation chain is self-contained in the sense required by the circularity analysis. Baseline estimation (DS-VIRE, Eqs. 1-7) fuses independent IMU, UWB, and infrared-marker observations; feature association (GP-SS) is a real engineering contribution; collaborative triangulation (Eqs. 8-12) produces metric sparse landmarks from multi-view observations; and the sparse-to-dense step (Eq. 13) fits four parameters (a, b, c, d) of an exponential functional form to those sparse landmarks, then applies the fitted transform to the entire monocular depth image. This is a fitted-parameter-to-dense-prediction scheme, but it is not circular: the prediction target is the dense depth map, which is not defined as the fitted landmarks, and the reported 2.3%-9.7% relative errors are computed in Table IX against external R3LIVE ground truth using the uCD metric (Eq. 17), not against the landmarks used in the fit. The exponential form is chosen based on an observed depth-distribution pattern (Fig. 7b) and is empirically benchmarked against linear and quadratic fitting; this is in-sample model selection rather than a circular reduction. No load-bearing uniqueness theorem is imported from the authors' prior work, and the self-citations appearing in the related-work section (e.g., [17], [53], [66]) are contextual and do not justify the central claim. The acknowledged communication-interruption limitation is a scope caveat, not a circular step. Any concerns about per-pixel generalization, unidirectional Chamfer metric, or the small number of evaluation frames are correctness-risk issues, not circularity.

Assumptions & free parameters 2 free parameters · 5 assumptions · 0 invented entities

The central claim rests on one fitted exponential mapping (4 parameters per frame), a condition-number threshold, and several domain assumptions about ground truth, formation, and monocular depth expressibility; no new physical entities are introduced.

free parameters (2)
  • exponential fitting parameters a, b, c, d (Eq. 13) = not reported; optimized per frame on 60-200 sparse landmarks via Ceres
    These four parameters map DepthAnythingV2's up-to-scale depth to metric depth across the whole dense image; the central mapping accuracy depends on them. Parameter c is redundant with a and b, so the model is overparameterized.
  • condition number threshold for landmark acceptance (Sec VI-A) = not reported
    Hand-chosen threshold discards triangulated landmarks; it affects which sparse points enter the exponential fit and the reported error.
assumptions (5)
  • domain assumption R3LIVE LiDAR-inertial reconstruction is accurate ground truth at 10-70 m depth in open outdoor scenes.
    Used in Sec VII-F to compute uCD and all reported dense mapping errors; no independent accuracy assessment of R3LIVE at these ranges is given.
  • domain assumption DepthAnythingV2's relative depth output is globally related to metric depth by a single exponential transform per frame.
    Sec VI-B and Eq (13) assume this; if false, the dense depth scaling will fail in depth regions or image areas not covered by sparse landmarks.
  • standard math PnP, IMU preintegration, and UWB residual models are unbiased and correctly fused in the sliding window.
    Sec IV-B Eqs (2)-(6) rely on standard estimation assumptions; no sensor noise covariances are reported.
  • ad hoc to paper Bidirectional yaw estimate Eq (7) treats a bright IR marker as a proxy for the side camera center and assumes near-hovering roll/pitch, with a level-frame correction.
    Sec IV-C; if the marker-camera offset is not negligible or the level correction is inaccurate, yaw estimates degrade. Yaw errors are shown in Fig 20 to be the dominant triangulation error source.
  • domain assumption The UAV formation stays within 5 m with low-latency WiFi; mapping pauses during communication interruptions.
    Sec VII-B1; the headline 70 m mapping result is demonstrated under this formation and communication envelope, not under arbitrary swarm geometry.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Flying Co-Stereo: Enabling Long-Range Aerial Dense Mapping via Collaborative Stereo Vision of Dynamic-Baseline." pith.science (2026). https://pith.science/paper/5NWY5ZCM

@misc{pith2026250600546,
  author       = {Pith},
  title        = {Pith review of: Flying Co-Stereo: Enabling Long-Range Aerial Dense Mapping via Collaborative Stereo Vision of Dynamic-Baseline},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/5NWY5ZCM}},
  note         = {Machine review of arXiv:2506.00546}
}
read the original abstract

Lightweight long-range mapping is critical for safe navigation of UAV swarms in large-scale unknown environments. Traditional stereo vision systems with fixed short baselines face limited perception ranges. To address this, we propose Flying Co-Stereo, a cross-agent collaborative stereo vision system that leverages the wide-baseline spatial configuration of two UAVs for long-range dense mapping. Key innovations include: (1) a dual-spectrum visual-inertial-ranging estimator for robust baseline estimation; (2) a hybrid feature association strategy combining deep learning-based cross-agent matching and optical-flow-based intra-agent tracking; (3) A sparse-to-dense depth recovery scheme,refining dense monocular depth predictions using exponential fitting of long-range triangulated sparse landmarks for precise metric-scale mapping. Experiments demonstrate the Flying Co-Stereo system achieves dense 3D mapping up to 70 meters with 2.3%-9.7% relative error, outperforming conventional systems by up to 350% in depth range and 450% in coverage area. The project webpage: https://xingxingzuo.github.io/flying_co_stereo

Figures

Figures reproduced from arXiv: 2506.00546 by the authors.

Figure 1
Figure 1. The system architecture of Flying Co-Stereo within our proposed Collaborative Dynamic-Baseline Stereo Mapping framework. The online baseline [PITH_FULL_IMAGE:figures/full_fig_p004_1.png] view at source ↗
Figure 2
Figure 2. The framework of the DS-VIRE algorithm. The DS-MVDT submodule [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figure 3
Figure 3. Pipeline of dual-spectrum marker-based visual detection and tracking. [PITH_FULL_IMAGE:figures/full_fig_p005_3.png] view at source ↗
Figures from the paper (20 more)
Figure 4
Figure 4. Figure 4: Illustration of bidirectional visual differential for relative yaw [PITH_FULL_IMAGE:figures/full_fig_p006_4.png]
Figure 5
Figure 5. Figure 5: The diagram of guidance-prediction cross-camera feature association. [PITH_FULL_IMAGE:figures/full_fig_p007_5.png]
Figure 6
Figure 6. Figure 6: The periodic matching guidance and continuous feature prediction. [PITH_FULL_IMAGE:figures/full_fig_p007_6.png]
Figure 8
Figure 8. Figure 8: The hardware layout of the custom-developed quadcopter UAV. [PITH_FULL_IMAGE:figures/full_fig_p009_8.png]
Figure 7
Figure 7. Figure 7: Comparison of different fitting methods for dense reconstruction. [PITH_FULL_IMAGE:figures/full_fig_p009_7.png]
Figure 9
Figure 9. Figure 9: The latency of meshed WiFi communication with different baselines. [PITH_FULL_IMAGE:figures/full_fig_p010_9.png]
Figure 11
Figure 11. Figure 11: Comparison of relative position and orientation estimates from the proposed DS-VIRE, Visual PnP, and VIO [PITH_FULL_IMAGE:figures/full_fig_p011_11.png]
Figure 12
Figure 12. Figure 12: Robustness evaluation of DS-MVDT in outdoor Field-A, featuring complex backgrounds and lighting disturbances. UAV trajectories are shown in [PITH_FULL_IMAGE:figures/full_fig_p012_12.png]
Figure 14
Figure 14. Figure 14: The visualization of cross-camera feature association in the real [PITH_FULL_IMAGE:figures/full_fig_p012_14.png]
Figure 13
Figure 13. Figure 13: Experiments of DS-MVDT with challenges from intense sunlight, [PITH_FULL_IMAGE:figures/full_fig_p012_13.png]
Figure 16
Figure 16. Figure 16: The number of the retained feature along consecutive frame sequence. [PITH_FULL_IMAGE:figures/full_fig_p013_16.png]
Figure 15
Figure 15. Figure 15: The run-time of different cross-camera feature association algorithms [PITH_FULL_IMAGE:figures/full_fig_p013_15.png]
Figure 17
Figure 17. Figure 17: Evaluation scenarios of the Flying Co-Stereo system in both real-world and simulated environments. [PITH_FULL_IMAGE:figures/full_fig_p014_17.png]
Figure 19
Figure 19. Figure 19: Comparison of the average condition number of landmarks triangu [PITH_FULL_IMAGE:figures/full_fig_p015_19.png]
Figure 18
Figure 18. Figure 18: The triangulation configurations for two UAVs flying in parallel. [PITH_FULL_IMAGE:figures/full_fig_p015_18.png]
Figure 21
Figure 21. Figure 21: The numerical analysis of landmark triangulation error with respect [PITH_FULL_IMAGE:figures/full_fig_p016_21.png]
Figure 22
Figure 22. Figure 22: The long-range dense reconstruction experiments in outdoor environments and photorealistic simulation. The visualization comprises seven columns: [PITH_FULL_IMAGE:figures/full_fig_p017_22.png]
Figure 23
Figure 23. Figure 23: Overall uCD comparison of different fitting methods across multiple [PITH_FULL_IMAGE:figures/full_fig_p017_23.png]
Figure 24
Figure 24. Figure 24: Comparison of dense reconstruction results from the sparse-to-dense pipeline of Flying Co-Stereo and a commercial stereo camera. The dense [PITH_FULL_IMAGE:figures/full_fig_p018_24.png]
Figure 25
Figure 25. Figure 25: Comparison of reconstruction from MVSAnywhere [57], SimpleRe [PITH_FULL_IMAGE:figures/full_fig_p018_25.png]

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

80 extracted references · 53 canonical work pages

  1. [1]

    Safety-assured high-speed navigation for mavs,

    Y . Ren, F. Zhu, G. Lu, Y . Cai, L. Yin, F. Kong, J. Lin, N. Chen, and F. Zhang, “Safety-assured high-speed navigation for mavs,”Science Robotics, vol. 10, no. 98, p. eado6187, 2025

    2025

  2. [2]

    Learning high-speed flight in the wild,

    A. Loquercio, E. Kaufmann, R. Ranftl, M. M ¨uller, V . Koltun, and D. Scaramuzza, “Learning high-speed flight in the wild,”Science Robotics, vol. 6, no. 59, p. eabg5810, 2021

    2021

  3. [3]

    Swarm of micro flying robots in the wild,

    X. Zhou, X. Wen, Z. Wang, Y . Gao, H. Li, Q. Wang, T. Yang, H. Lu, Y . Cao, C. Xu, and F. Gao, “Swarm of micro flying robots in the wild,” Science Robotics, vol. 7, no. 66, p. eabm5954, 2022

    work page 2022

  4. [4]

    Search and rescue under the forest canopy using multiple uavs,

    Y . Tian, K. Liu, K. Ok, L. Tran, D. Allen, N. Roy, and J. P. How, “Search and rescue under the forest canopy using multiple uavs,”The International Journal of Robotics Research, vol. 39, no. 10-11, pp. 1201– 1221, 2020

    2020

  5. [5]

    Lic-fusion: Lidar- inertial-camera odometry,

    X. Zuo, P. Geneva, W. Lee, Y . Liu, and G. Huang, “Lic-fusion: Lidar- inertial-camera odometry,” inProc. IEEE/RSJ International Conference on Intelligent Robots and Systems, Macau, China, 2019, pp. 5848–5854

    work page 2019

  6. [6]

    LIC-Fusion 2.0: LiDAR-inertial-camera odometry with sliding-window plane-feature tracking,

    X. Zuo, Y . Yang, P. Geneva, J. Lv, Y . Liu, G. Huang, and M. Pollefeys, “LIC-Fusion 2.0: LiDAR-inertial-camera odometry with sliding-window plane-feature tracking,” inProc. IEEE/RSJ International Conference on Intelligent Robots and Systems, Las Vegas, NV , USA, 2020, pp. 5112– 5119

    work page 2020

  7. [7]

    Clins: Continuous- time trajectory estimation for lidar-inertial system,

    J. Lv, K. Hu, J. Xu, Y . Liu, X. Ma, and X. Zuo, “Clins: Continuous- time trajectory estimation for lidar-inertial system,” in2021 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS). IEEE, 2021, pp. 6657–6663

    work page 2021

  8. [8]

    Viral-fusion: A visual-inertial-ranging-lidar sensor fusion approach,

    T.-M. Nguyen, M. Cao, S. Yuan, Y . Lyu, T. H. Nguyen, and L. Xie, “Viral-fusion: A visual-inertial-ranging-lidar sensor fusion approach,” IEEE Transactions on Robotics, vol. 38, no. 2, pp. 958–977, 2021

    work page 2021

Show all 80 references
  1. [9]

    Coco-LIC: Continuous-time tightly-coupled lidar-inertial-camera odometry using non-uniform b-spline,

    X. Lang, C. Chen, K. Tang, Y . Ma, J. Lv, Y . Liu, and X. Zuo, “Coco-LIC: Continuous-time tightly-coupled lidar-inertial-camera odometry using non-uniform b-spline,”IEEE Robotics and Automation Letters, vol. 8, no. 11, pp. 7074–7081, 2023

    2023

  2. [10]

    Gaussian-LIC: Real-time photo-realistic slam with gaussian splatting and lidar-inertial-camera fusion,

    X. Lang, L. Li, C. Wu, C. Zhao, L. Liu, Y . Liu, J. Lv, and X. Zuo, “Gaussian-LIC: Real-time photo-realistic slam with gaussian splatting and lidar-inertial-camera fusion,” in2025 IEEE International Conference on Robotics and Automation (ICRA). IEEE, 2025

    2025

  3. [11]

    Computa- tionally efficient trajectory planning for high speed obstacle avoidance of a quadrotor with active sensing,

    G. Chen, D. Sun, W. Dong, X. Sheng, X. Zhu, and H. Ding, “Computa- tionally efficient trajectory planning for high speed obstacle avoidance of a quadrotor with active sensing,”IEEE Robotics and Automation Letters, vol. 6, no. 2, pp. 3365–3372, 2021

    2021

  4. [12]

    An active sense and avoid system for flying robots in dynamic environments,

    G. Chen, W. Dong, X. Sheng, X. Zhu, and H. Ding, “An active sense and avoid system for flying robots in dynamic environments,”IEEE/ASME Transactions on Mechatronics, vol. 26, no. 2, pp. 668–678, 2021

    2021

  5. [13]

    Flexible trinocular: Non-rigid multi-camera-imu dense reconstruction for uav navigation and mapping,

    T. Hinzmann, C. Cadena, J. Nieto, and R. Siegwart, “Flexible trinocular: Non-rigid multi-camera-imu dense reconstruction for uav navigation and mapping,” in2019 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS). IEEE, 2019, pp. 1137–1142

    2019

  6. [14]

    Variable baseline/resolution stereo,

    D. Gallup, J.-M. Frahm, P. Mordohai, and M. Pollefeys, “Variable baseline/resolution stereo,” in2008 IEEE conference on computer vision and pattern recognition. IEEE, 2008, pp. 1–8

    2008

  7. [15]

    Collaborative stereo,

    M. W. Achtelik, S. Weiss, M. Chli, F. Dellaerty, and R. Siegwart, “Collaborative stereo,” in2011 IEEE/RSJ International Conference on Intelligent Robots and Systems. IEEE, 2011, pp. 2242–2248

    2011

  8. [16]

    Distributed variable-baseline stereo slam from two uavs,

    M. Karrer and M. Chli, “Distributed variable-baseline stereo slam from two uavs,” in2021 IEEE International Conference on Robotics and Automation (ICRA). IEEE, 2021, pp. 82–88

    2021

  9. [17]

    A collaborative stereo camera with two uavs for long-distance mapping of urban buildings,

    Z. Wang and W. Dong, “A collaborative stereo camera with two uavs for long-distance mapping of urban buildings,” in2024 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS). IEEE, 2024, pp. 7944–7951

    2024

  10. [18]

    Agile Formation Control of Drone Flocking Enhanced With Active Vision-Based Relative Localiza- tion,

    P. Zhang, G. Chen, Y . Li, and W. Dong, “Agile Formation Control of Drone Flocking Enhanced With Active Vision-Based Relative Localiza- tion,”IEEE Robotics and Automation Letters, vol. 7, no. 3, pp. 6359– 6366, Jul. 2022

    2022

  11. [19]

    Fiducial markers for pose estimation: Overview, appli- cations and experimental comparison of the artag, apriltag, aruco and stag markers,

    M. Kalaitzakis, B. Cain, S. Carroll, A. Ambrosi, C. Whitehead, and N. Vitzilaios, “Fiducial markers for pose estimation: Overview, appli- cations and experimental comparison of the artag, apriltag, aruco and stag markers,”Journal of Intelligent&Robotic Systems, vol. 101, no. 4...

    2021

  12. [20]

    Vision-based drone flocking in outdoor environments,

    F. Schilling, F. Schiano, and D. Floreano, “Vision-based drone flocking in outdoor environments,”IEEE Robotics and Automation Letters, vol. 6, no. 2, pp. 2954–2961, 2021

    2021

  13. [21]

    Vision-based multi-mav localization with anonymous relative measurements using coupled probabilistic data association filter,

    T. Nguyen, K. Mohta, C. J. Taylor, and V . Kumar, “Vision-based multi-mav localization with anonymous relative measurements using coupled probabilistic data association filter,” in2020 IEEE International Conference on Robotics and Automation (ICRA). IEEE, 2020, pp. 3349–3355

    2020

  14. [22]

    Tracking and Relative Localization of Drone Swarms With a Vision- Based Headset,

    M. Pavliv, F. Schiano, C. Reardon, D. Floreano, and G. Loianno, “Tracking and Relative Localization of Drone Swarms With a Vision- Based Headset,”IEEE Robotics and Automation Letters, vol. 6, no. 2, pp. 1455–1462, 2021

    2021

  15. [23]

    6- dof object pose from semantic keypoints,

    G. Pavlakos, X. Zhou, A. Chan, K. G. Derpanis, and K. Daniilidis, “6- dof object pose from semantic keypoints,” in2017 IEEE international conference on robotics and automation (ICRA). IEEE, 2017, pp. 2011– 2018

    2017

  16. [24]

    Vi-rpe: Visual-inertial relative pose estimation for aerial vehicles,

    L. Teixeira, F. Maffra, M. Moos, and M. Chli, “Vi-rpe: Visual-inertial relative pose estimation for aerial vehicles,”IEEE Robotics and Automa- tion Letters, vol. 3, no. 4, pp. 2770–2777, 2018

    2018

  17. [25]

    Robust Visual Positioning of the UA V for the Under Bridge Inspection With a Ground Guided Vehicle,

    Z. Wang, S. Liu, G. Chen, and W. Dong, “Robust Visual Positioning of the UA V for the Under Bridge Inspection With a Ground Guided Vehicle,”IEEE Transactions on Instrumentation and Measurement, vol. 71, pp. 1–10, 2022

    2022

  18. [26]

    Uvdar system for visual relative localization with application to leader–follower formations of multirotor uavs,

    V . Walter, N. Staub, A. Franchi, and M. Saska, “Uvdar system for visual relative localization with application to leader–follower formations of multirotor uavs,”IEEE Robotics and Automation Letters, vol. 4, no. 3, pp. 2637–2644, 2019

    2019

  19. [27]

    Crepes: Cooperative relative pose estimation system,

    Z. Xun, J. Huang, Z. Li, Z. Ying, Y . Wang, C. Xu, F. Gao, and Y . Cao, “Crepes: Cooperative relative pose estimation system,” in2023 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS). IEEE, 2023, pp. 5274–5281

    2023

  20. [28]

    Decentralized visual- inertial-uwb fusion for relative state estimation of aerial swarm,

    H. Xu, L. Wang, Y . Zhang, K. Qiu, and S. Shen, “Decentralized visual- inertial-uwb fusion for relative state estimation of aerial swarm,” in 2020 IEEE international conference on robotics and automation (ICRA). IEEE, 2020, pp. 8776–8782

    2020

  21. [29]

    Omni-swarm: A decentralized omnidirectional visual–inertial– uwb state estimation system for aerial swarms,

    H. Xu, Y . Zhang, B. Zhou, L. Wang, X. Yao, G. Meng, and S. Shen, “Omni-swarm: A decentralized omnidirectional visual–inertial– uwb state estimation system for aerial swarms,”IEEE Transactions on Robotics, vol. 38, no. 6, pp. 3374–3394, 2022. 20

    2022

  22. [30]

    D2slam: Decentralized and distributed collaborative visual-inertial slam system for aerial swarm,

    H. Xu, P. Liu, X. Chen, and S. Shen, “D2slam: Decentralized and distributed collaborative visual-inertial slam system for aerial swarm,” IEEE Transactions on Robotics, 2024

    2024

  23. [31]

    Stereo processing by semiglobal matching and mu- tual information,

    H. Hirschmuller, “Stereo processing by semiglobal matching and mu- tual information,”IEEE Transactions on pattern analysis and machine intelligence, vol. 30, no. 2, pp. 328–341, 2007

    2007

  24. [32]

    Efficient large-scale stereo matching,

    A. Geiger, M. Roser, and R. Urtasun, “Efficient large-scale stereo matching,” inAsian conference on computer vision. Springer, 2010, pp. 25–38

    2010

  25. [33]

    Coslam: Collaborative visual slam in dynamic environments,

    D. Zou and P. Tan, “Coslam: Collaborative visual slam in dynamic environments,”IEEE transactions on pattern analysis and machine intelligence, vol. 35, no. 2, pp. 354–366, 2012

    2012

  26. [34]

    Cvi-slam—collaborative visual- inertial slam,

    M. Karrer, P. Schmuck, and M. Chli, “Cvi-slam—collaborative visual- inertial slam,”IEEE Robotics and Automation Letters, vol. 3, no. 4, pp. 2762–2769, 2018

    2018

  27. [35]

    Ccm-slam: Robust and efficient centralized collaborative monocular simultaneous localization and mapping for robotic teams,

    P. Schmuck and M. Chli, “Ccm-slam: Robust and efficient centralized collaborative monocular simultaneous localization and mapping for robotic teams,”Journal of Field Robotics, vol. 36, no. 4, pp. 763–781, 2019

    2019

  28. [36]

    Covins: Visual-inertial slam for centralized collaboration,

    P. Schmuck, T. Ziegler, M. Karrer, J. Perraudin, and M. Chli, “Covins: Visual-inertial slam for centralized collaboration,” in2021 IEEE Inter- national Symposium on Mixed and Augmented Reality Adjunct (ISMAR- Adjunct). IEEE, 2021, pp. 171–176

    2021

  29. [37]

    ORB-SLAM2: An Open-Source SLAM System for Monocular, Stereo, and RGB-D Cameras,

    R. Mur-Artal and J. D. Tard ´os, “ORB-SLAM2: An Open-Source SLAM System for Monocular, Stereo, and RGB-D Cameras,”IEEE Transac- tions on Robotics, vol. 33, no. 5, pp. 1255–1262, Oct. 2017

    2017

  30. [38]

    Distinctive image features from scale-invariant keypoints,

    L. David, “Distinctive image features from scale-invariant keypoints,” International journal of computer vision, vol. 60, pp. 91–110, 2004

    2004

  31. [39]

    Surf: Speeded up robust features,

    H. Bay, T. Tuytelaars, and L. Van Gool, “Surf: Speeded up robust features,” inComputer Vision–ECCV 2006: 9th European Conference on Computer Vision. Springer, 2006, pp. 404–417

    2006

  32. [40]

    Superpoint: Self- supervised interest point detection and description,

    D. DeTone, T. Malisiewicz, and A. Rabinovich, “Superpoint: Self- supervised interest point detection and description,” inProceedings of the IEEE conference on computer vision and pattern recognition workshops, 2018, pp. 224–236

    2018

  33. [41]

    Superglue: Learning feature matching with graph neural networks,

    P.-E. Sarlin, D. DeTone, T. Malisiewicz, and A. Rabinovich, “Superglue: Learning feature matching with graph neural networks,” inProceedings of the IEEE/CVF conference on computer vision and pattern recognition, 2020, pp. 4938–4947

    2020

  34. [42]

    Lightglue: Local feature matching at light speed,

    P. Lindenberger, P.-E. Sarlin, and M. Pollefeys, “Lightglue: Local feature matching at light speed,” inProceedings of the IEEE/CVF International Conference on Computer Vision, 2023, pp. 17 627–17 638

    2023

  35. [43]

    Loftr: Detector- free local feature matching with transformers,

    J. Sun, Z. Shen, Y . Wang, H. Bao, and X. Zhou, “Loftr: Detector- free local feature matching with transformers,” inProceedings of the IEEE/CVF conference on computer vision and pattern recognition, 2021, pp. 8922–8931

    2021

  36. [44]

    Distributed visual-inertial cooperative localization,

    P. Zhu, P. Geneva, W. Ren, and G. Huang, “Distributed visual-inertial cooperative localization,” in2021 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS). IEEE, 2021, pp. 8714–8721

    2021

  37. [45]

    Svo: Fast semi-direct monocular visual odometry,

    C. Forster, M. Pizzoli, and D. Scaramuzza, “Svo: Fast semi-direct monocular visual odometry,” in2014 IEEE international conference on robotics and automation (ICRA). IEEE, 2014, pp. 15–22

    2014

  38. [46]

    Remode: Probabilistic, monocular dense reconstruction in real time,

    M. Pizzoli, C. Forster, and D. Scaramuzza, “Remode: Probabilistic, monocular dense reconstruction in real time,” in2014 IEEE international conference on robotics and automation (ICRA). IEEE, 2014, pp. 2609– 2616

    2014

  39. [47]

    Lsd-slam: Large-scale direct monocular slam,

    J. Engel, T. Sch ¨ops, and D. Cremers, “Lsd-slam: Large-scale direct monocular slam,” inEuropean conference on computer vision. Springer, 2014, pp. 834–849

    2014

  40. [48]

    Kimera: an open- source library for real-time metric-semantic localization and mapping,

    A. Rosinol, M. Abate, Y . Chang, and L. Carlone, “Kimera: an open- source library for real-time metric-semantic localization and mapping,” in2020 IEEE International Conference on Robotics and Automation (ICRA). IEEE, 2020, pp. 1689–1696

    2020

  41. [49]

    Real-time mesh-based scene estimation for aerial inspection,

    L. Teixeira and M. Chli, “Real-time mesh-based scene estimation for aerial inspection,” in2016 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS). IEEE, 2016, pp. 4863–4869

    2016

  42. [50]

    Cnn-slam: Real-time dense monocular slam with learned depth prediction,

    K. Tateno, F. Tombari, I. Laina, and N. Navab, “Cnn-slam: Real-time dense monocular slam with learned depth prediction,” inProceedings of the IEEE conference on computer vision and pattern recognition, 2017, pp. 6243–6252

    2017

  43. [51]

    Codeslam—learning a compact, optimisable representation for dense visual slam,

    M. Bloesch, J. Czarnowski, R. Clark, S. Leutenegger, and A. J. Davison, “Codeslam—learning a compact, optimisable representation for dense visual slam,” inProceedings of the IEEE conference on computer vision and pattern recognition, 2018, pp. 2560–2568

    2018

  44. [52]

    Codemap- ping: Real-time dense mapping for sparse slam using compact scene representations,

    H. Matsuki, R. Scona, J. Czarnowski, and A. J. Davison, “Codemap- ping: Real-time dense mapping for sparse slam using compact scene representations,”IEEE Robotics and Automation Letters, vol. 6, no. 4, pp. 7105–7112, 2021

    2021

  45. [53]

    Codevio: Visual-inertial odometry with learned optimizable dense depth,

    X. Zuo, N. Merrill, W. Li, Y . Liu, M. Pollefeys, and G. Huang, “Codevio: Visual-inertial odometry with learned optimizable dense depth,” in2021 IEEE International Conference on Robotics and Automation (ICRA). IEEE, 2021, pp. 14 382–14 388

    2021

  46. [54]

    Simplerecon: 3d reconstruction without 3d convolutions,

    M. Sayed, J. Gibson, J. Watson, V . Prisacariu, M. Firman, and C. Go- dard, “Simplerecon: 3d reconstruction without 3d convolutions,” in European Conference on Computer Vision. Springer, 2022, pp. 1–19

    2022

  47. [55]

    Simplemapping: Real- time visual-inertial dense mapping with deep multi-view stereo,

    Y . Xin, X. Zuo, D. Lu, and S. Leutenegger, “Simplemapping: Real- time visual-inertial dense mapping with deep multi-view stereo,” in 2023 IEEE International Symposium on Mixed and Augmented Reality (ISMAR). IEEE, 2023, pp. 273–282

    2023

  48. [56]

    Scannet: Richly-annotated 3d reconstructions of indoor scenes,

    A. Dai, A. X. Chang, M. Savva, M. Halber, T. Funkhouser, and M. Nießner, “Scannet: Richly-annotated 3d reconstructions of indoor scenes,” inProceedings of the IEEE conference on computer vision and pattern recognition, 2017, pp. 5828–5839

    2017

  49. [57]

    MVSAnywhere: Zero shot multi-view stereo,

    S. Izquierdo, M. Sayed, M. Firman, G. Garcia-Hernando, D. Tur- mukhambetov, J. Civera, O. Mac Aodha, G. J. Brostow, and J. Watson, “MVSAnywhere: Zero shot multi-view stereo,” inCVPR, 2025

    2025

  50. [58]

    Sparse sensing for resource-constrained depth reconstruction,

    F. Ma, L. Carlone, U. Ayaz, and S. Karaman, “Sparse sensing for resource-constrained depth reconstruction,” in2016 IEEE/RSJ Interna- tional Conference on Intelligent Robots and Systems (IROS). IEEE, 2016, pp. 96–103

    2016

  51. [60]

    Aerial single- view depth completion with image-guided uncertainty estimation,

    L. Teixeira, M. R. Oswald, M. Pollefeys, and M. Chli, “Aerial single- view depth completion with image-guided uncertainty estimation,”IEEE Robotics and Automation Letters, vol. 5, no. 2, pp. 1055–1062, 2020

    2020

  52. [61]

    Unsupervised depth com- pletion from visual inertial odometry,

    A. Wong, X. Fei, S. Tsuei, and S. Soatto, “Unsupervised depth com- pletion from visual inertial odometry,”IEEE Robotics and Automation Letters, vol. 5, no. 2, pp. 1899–1906, 2020

    1906

  53. [62]

    Indoor segmentation and support inference from rgbd images,

    N. Silberman, D. Hoiem, P. Kohli, and R. Fergus, “Indoor segmentation and support inference from rgbd images,” inComputer Vision–ECCV 2012: 12th European Conference on Computer Vision, Florence, Italy, October 7-13, 2012, Proceedings, Part V 12. Springer, 2012, pp. 746– 760

    2012

  54. [63]

    Vision meets robotics: The kitti dataset,

    A. Geiger, P. Lenz, C. Stiller, and R. Urtasun, “Vision meets robotics: The kitti dataset,”International Journal of Robotics Research (IJRR), 2013

    2013

  55. [64]

    Towards robust monocular depth estimation: Mixing datasets for zero-shot cross- dataset transfer,

    R. Ranftl, K. Lasinger, D. Hafner, K. Schindler, and V . Koltun, “Towards robust monocular depth estimation: Mixing datasets for zero-shot cross- dataset transfer,”IEEE transactions on pattern analysis and machine intelligence, vol. 44, no. 3, pp. 1623–1637, 2020

    2020

  56. [65]

    Depth anything v2,

    L. Yang, B. Kang, Z. Huang, Z. Zhao, X. Xu, J. Feng, and H. Zhao, “Depth anything v2,”Advances in Neural Information Processing Sys- tems, vol. 37, pp. 21 875–21 911, 2024

    2024

  57. [66]

    Nerf-vo: Real-time sparse visual odometry with neural radiance fields,

    J. Naumann, B. Xu, S. Leutenegger, and X. Zuo, “Nerf-vo: Real-time sparse visual odometry with neural radiance fields,”IEEE Robotics and Automation Letters, 2024

    2024

  58. [67]

    Vision transformers for dense prediction,

    R. Ranftl, A. Bochkovskiy, and V . Koltun, “Vision transformers for dense prediction,” inProceedings of the IEEE/CVF international conference on computer vision, 2021, pp. 12 179–12 188

    2021

  59. [68]

    Visual-inertial slam as simple as a, b, vins,

    N. Merrill and G. Huang, “Visual-inertial slam as simple as a, b, vins,” arXiv preprint arXiv:2406.05969, 2024

    2024 arXiv

  60. [69]

    V oxblox: Incremental 3d euclidean signed distance fields for on-board mav planning,

    H. Oleynikova, Z. Taylor, M. Fehr, R. Siegwart, and J. Nieto, “V oxblox: Incremental 3d euclidean signed distance fields for on-board mav planning,” in2017 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS). IEEE, 2017, pp. 1366–1373

    2017

  61. [70]

    Openvins: A research platform for visual-inertial estimation,

    P. Geneva, K. Eckenhoff, W. Lee, Y . Yang, and G. Huang, “Openvins: A research platform for visual-inertial estimation,” in2020 IEEE Inter- national Conference on Robotics and Automation (ICRA). IEEE, 2020, pp. 4666–4672

    2020

  62. [71]

    Yolov4: Op- timal speed and accuracy of object detection,

    A. Bochkovskiy, C.-Y . Wang, and H.-Y . M. Liao, “Yolov4: Op- timal speed and accuracy of object detection,”arXiv preprint arXiv:2004.10934, 2020

    2004 arXiv

  63. [72]

    Faster and better: A machine learning approach to corner detection,

    E. Rosten, R. Porter, and T. Drummond, “Faster and better: A machine learning approach to corner detection,”IEEE transactions on pattern analysis and machine intelligence, vol. 32, no. 1, pp. 105–119, 2008

    2008

  64. [73]

    An iterative image registration technique with an application to stereo vision,

    B. D. Lucas and T. Kanade, “An iterative image registration technique with an application to stereo vision,” inIJCAI’81: 7th international joint conference on Artificial intelligence, vol. 2, 1981, pp. 674–679

    1981

  65. [74]

    Infinitesimal plane-based pose estimation,

    T. Collins and A. Bartoli, “Infinitesimal plane-based pose estimation,” International journal of computer vision, vol. 109, no. 3, pp. 252–286, 2014. 21

    2014

  66. [75]

    Ceres solver: Tutorial & reference,

    S. Agarwal, K. Mierleet al., “Ceres solver: Tutorial & reference,” Google Inc, vol. 2, no. 72, p. 8, 2012

    2012

  67. [76]

    Unsupervised depth completion with calibrated backprojection layers,

    A. Wong and S. Soatto, “Unsupervised depth completion with calibrated backprojection layers,” inProceedings of the IEEE/CVF International Conference on Computer Vision, 2021, pp. 12 747–12 756

    2021

  68. [77]

    Visual inertial odometry swarm: An autonomous swarm of vision-based quadrotors,

    A. Weinstein, A. Cho, G. Loianno, and V . Kumar, “Visual inertial odometry swarm: An autonomous swarm of vision-based quadrotors,” IEEE Robotics and Automation Letters, vol. 3, no. 3, pp. 1801–1807, 2018

    2018

  69. [78]

    Visual object tracking using adaptive correlation filters,

    D. S. Bolme, J. R. Beveridge, B. A. Draper, and Y . M. Lui, “Visual object tracking using adaptive correlation filters,” in2010 IEEE computer society conference on computer vision and pattern recognition. IEEE, 2010, pp. 2544–2550

    2010

  70. [79]

    Random sample consensus: a paradigm for model fitting with applications to image analysis and automated cartography,

    M. A. Fischler and R. C. Bolles, “Random sample consensus: a paradigm for model fitting with applications to image analysis and automated cartography,”Commun. ACM, vol. 24, no. 6, p. 381–395, Jun. 1981

    1981

  71. [80]

    Airvo: An illumination- robust point-line visual odometry,

    K. Xu, Y . Hao, S. Yuan, C. Wang, and L. Xie, “Airvo: An illumination- robust point-line visual odometry,” in2023 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS). IEEE, 2023, pp. 3429–3436

    2023

  72. [81]

    R3live: A robust, real-time, rgb-colored, lidar- inertial-visual tightly-coupled state estimation and mapping package,

    J. Lin and F. Zhang, “R3live: A robust, real-time, rgb-colored, lidar- inertial-visual tightly-coupled state estimation and mapping package,” in2022 International Conference on Robotics and Automation (ICRA). IEEE, 2022, pp. 10 672–10 678

    2022

Pith tools

Reviewed August 7, 2026 · model on record in the stance chip above.