Pith. sign in

REVIEW 4 major objections 4 minor 34 references

Differentiable Biomechanics for Markerless Motion Capture in Upper Limb Stroke Rehabilitation: A Comparison with Optical Motion Capture

T0 review · 4 major / 4 minor · reviewed 2026-08-12 · deepseek-v4-flash

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

Pith's one-line read This paper claims that markerless motion capture from five synchronized webcams, combined with differentiable biomechanical optimization, recovers upper-limb drinking-task kinematics in stroke patients whose trajectories agree with…

desk verdict Real first validation of end-to-end MMC against OMC in stroke, but the abstract overstates absolute accuracy by hiding per-trial bias removal. read the letter →

arxiv 2411.14992 v1 pith:GZKOKHFR submitted 2024-11-22 cs.CV

Tim Unger , Arash Sal Moslehian , J.D. Peiffer , Johann Ullrich , Roger Gassert , Olivier Lambercy , R. James Cotton , Chris Awai Easthope This is my paper · ORCID
classification cs.CV
keywords markerlessmotioncaptureopticaldifferentiablebiomechanicsstrokerehabilitationdrinkingtaskupperlimbkinematicsinversemovementqualitymeasures
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

This paper claims that a markerless motion capture (MMC) pipeline built on differentiable biomechanics can recover upper-limb kinematics in stroke patients with accuracy approaching that of marker-based optical motion capture (OMC). Fifteen stroke patients performed the standardized drinking task while five synchronized webcams recorded them alongside 3D marker cameras; the two systems' kinematic trajectories were then compared. After removing each trial's static offset and optimizing a small time shift, the paper reports median correlations above 0.95 for most joint-angle, trunk, and velocity trajectories, with joint-angle root-mean-square errors of roughly 2–5 degrees. If the claim holds, clinics could obtain quantitative movement-quality measures without marker placement, expensive equipment, or extensive setup, using only consumer webcams. The central requirement for interpreting the result is that the removed static offsets are harmless nuisance parameters rather than real differences in joint-angle definition between the two modeling pipelines.

What carries the argument

The load-bearing mechanism is an end-to-end differentiable biomechanical optimization that replaces the two-stage pipeline of marker trajectories followed by inverse kinematics with a single learnable mapping. An implicit neural function $f_\phi: t \to \theta$ maps time to joint angles, and the parameters $\phi$, together with per-participant body scaling and marker-offset parameters, are optimized by projecting a forward-kinematic model of the upper limb into calibrated webcam views and minimizing reprojection error. A bilevel optimization passes body-scale parameters across multiple trials, so one participant's scale is learned jointly with all their trajectory reconstructions. Before comparing outputs, the paper removes a per-trial static bias and optimizes a time lag up to 0.25 seconds, which is why the reported agreement describes dynamic waveform shape rather than absolute joint angles.

What would settle it

Recompute all agreement metrics without subtracting the per-trial mean difference, in particular reporting shoulder-flexion RMSE between MMC and OMC before bias removal; if that error stands near the reported median bias of about 23.5 degrees rather than near 2–3 degrees, the claimed accuracy depends entirely on treating a large systematic difference as removable noise. A complementary test is to record the same participants with dual-plane fluoroscopy alongside both systems and check whether MMC's shoulder-flexion waveform tracks true bone motion as closely as it tracks marker-based angles only after offset removal.

Watch

Extended reading notes

Core claim

The study's central claim is that an end-to-end differentiable markerless motion capture approach is nearing the accuracy of marker-based optical motion capture for measuring upper limb kinematics in stroke patients performing a drinking task. Instead of the conventional two-stage OMC workflow—3D marker trajectories followed by inverse kinematics on a scaled model—the MMC pipeline optimizes a neural implicit function that maps time to joint angles, jointly with body scale and marker offset parameters, by projecting a physics-based biomechanical model into five webcam views. The authors find median Pearson correlation r > 0.95 for a majority of kinematic trajectories and median RMSE values of 2–5 degrees for joint angles, 0.04 m/s for end-effector velocity, and 6 mm for trunk displacement after removing per-trial static bias and aligning time lag. They report that trial-to-trial biases between systems were consistent within participant sessions, with interquartile ranges of about 1–3 degrees for joint angles, 0.01 m/s, and 3 mm, and that reconstruction failed in only 1.6% of trials. The conclusion is that MMC for arm tracking is approaching marker-based accuracy and supports potential use in clinical settings.

Load-bearing premise

The load-bearing premise is that the static offset and time lag removed before computing every agreement metric are mere calibration nuisances, not genuine differences in how the two systems define or scale joint angles; if those offsets reflect real biomechanical disagreement, the reported accuracy is overstated.

Editorial extensions

If this is right

  • With per-session bias correction, webcam-based MMC could replace marker-based systems for extracting drinking-task movement quality measures in clinical stroke assessments, eliminating marker placement and reducing cost and setup time.
  • The low and stable trial-to-trial bias implies that a participant-specific offset could be estimated from a single calibration trial and applied to later sessions, enabling longitudinal monitoring of recovery.
  • Because the reported errors fall within the session-to-session variability previously observed for OMC itself, MMC-based measures may be as clinically interpretable as marker-based ones for patients with mild to moderate impairment.
  • Metrics such as number of movement units and interjoint coordination need refinement or replacement before smoothness and coordination scores are fully comparable across systems.
  • Averaging multiple trials, as already recommended for OMC protocols, also improves MMC correlations, so clinical practice could retain multi-trial designs.

Reading between the lines

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

  • A decisive test the paper does not run is to compare both OMC and MMC against dual-plane fluoroscopy of bone motion; until then, the reported 'accuracy' is agreement with a marker-based system that itself carries marker-registration and soft-tissue errors.
  • Because the largest removed bias reaches about 23.5 degrees for shoulder flexion, the headline 2–5 degree RMSE should be read as waveform-shape agreement after offset removal; absolute MMC joint angles are not yet interchangeable with OMC values without a per-session calibration.
  • The pipeline's reliance on 87 detected keypoints and five calibrated cameras leaves room to test whether monocular or fewer-camera configurations preserve the reported accuracy in everyday clinical environments.
  • Extending the same differentiable machinery to finger degrees of freedom and to other functional tasks is a natural next step, since the current models lock leg and finger motion.
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 / 4 minor

Summary. This manuscript compares a markerless motion capture (MMC) pipeline based on differentiable biomechanics (Cotton et al.'s end-to-end approach) against marker-based optical motion capture (OMC) in 15 stroke patients performing a standardized drinking task. The authors report high agreement between the two systems after per-trial bias removal and time-lag optimization, with median correlations above 0.95 for most kinematic trajectories and RMSE values of 2–5 degrees for joint angles, and conclude that MMC is 'approaching the accuracy of marker-based methods' and could be used in clinical settings. The paper includes a large dataset (1160 trials), a clinically relevant task, and open-source analysis code.

Significance. If the central claim were fully supported, this would be a valuable contribution: it is the first comparison of the end-to-end differentiable MMC approach against OMC in neurological patients, uses a clinically recommended task, and provides reproducible code. The paper also honestly reports several limitations, including OMC phase classification and model differences. However, the load-bearing conclusion rests on agreement metrics computed after subtracting per-trial static bias and optimizing time lag. Because the removed shoulder-flexion bias reaches about 23.5 degrees, the reported RMSE and correlation values do not by themselves establish that MMC can measure absolute joint angles or movement quality measures without an OMC-based calibration. The study's strengths are the scale of the dataset, the clinical relevance, and the transparency about many methodological choices, but the interpretation of the headline accuracy numbers needs substantial revision.

major comments (4)
  1. [II.D.1, Table I, Abstract] The abstract states that 'median RMSE values ranging from 2-5 degrees for joint angles' were observed, without mentioning that these RMSEs are computed after subtracting a per-trial static bias and optimizing a per-trial time lag of up to 0.25 s. Table I shows a median bias of -23.55 degrees for shoulder flexion and -7.52 degrees for elbow extension. Removing these offsets before computing RMSE makes the reported values measures of waveform similarity after alignment, not absolute joint-angle accuracy. The conclusion that MMC is 'approaching the accuracy of marker-based methods' is therefore not directly supported. The authors should report absolute (offset-unadjusted) RMSE and bias, and discuss whether a standalone MMC deployment could estimate and remove such offsets without an OMC reference. If the large offsets reflect genuine differences in joint-angle definitions or model scaling between OpenSim and MuJoCo, then the offset-corrected RMSE understates the systematic disagreement by tens of degrees.
  2. [II.D.2, III.C.2] Phase classification for both systems is derived exclusively from OMC end-effector velocity. This guarantees that phase boundaries and phase durations are identical for MMC and OMC, which inflates agreement for temporal measures such as time to peak velocity (rs = 0.98, rav = 1.00). The manuscript acknowledges this limitation but still presents these temporal agreement values as evidence of MMC accuracy. The authors should either validate phase classification using MMC alone or clearly re-label these results as reflecting kinematic agreement under OMC-defined phases, and temper the corresponding claims about temporal accuracy.
  3. [II.D.1, II.D.2] It is unclear whether the per-trial bias removal described in Section II.D.1 was also applied to the kinematic trajectories before computing the movement quality measures in Section II.D.2. If the bias is removed before extracting maxima such as shoulder flexion and elbow extension, then the high correlations in Table III (e.g., rs = 0.97 and 0.98) are correlations of offset-corrected waveforms, not agreement of absolute peak values. The manuscript should state explicitly which trajectory version (raw or bias-removed) was used for each movement quality measure, and if bias removal was applied, report the absolute values and their differences.
  4. [IV, Discussion] The statement that 'the differences in kinematic trajectories between OMC and MMC fall within the session-to-session variability observed with OMC alone [13]' is not supported by the cited reference. Reference [13] (Uchida and Seth) quantifies marker registration and model scaling uncertainty in inverse analyses, not session-to-session repeatability of OMC in a clinical population. The authors should either provide a direct comparison with an actual OMC test-retest study or revise the claim.
minor comments (4)
  1. [II.C.1] There is a typo in 'biomechancical model' in the first paragraph of Section II.C.1.
  2. [References] Reference [7] contains a typo: 'compensationg' should be 'compensation'.
  3. [Table I] The table would be easier to read if the rows for each kinematic variable were visually separated or grouped, as the current dense layout makes cross-variable comparisons difficult.
  4. [II.D.1] The description of the time-lag optimization says 'shifting one kinematic trajectory over the other, up to 0.25 seconds' but does not specify the direction of the shift or the interpolation method used for sub-sample shifts; adding this detail would improve reproducibility.

Circularity Check

1 steps flagged · score 3.0 of 10

Headline RMSEs are post-fit residuals after per-trial bias removal (shoulder-flexion bias ≈23.6°), so the 'approaching OMC accuracy' claim is partly a function of the comparison's own preprocessing; the core waveform comparison is nonetheless externally benchmarked.

  1. fitted input called prediction [Section II.D.1 (Kinematic Trajectories); Table I; Abstract]
    "For each trial and kinematic trajectory, we calculated the static offset between the two systems as the difference in the means of MMC and OMC trajectory. This bias (offset) was then added to the OMC signal. ... All further analyses of noise were conducted with bias-removed kinematic trajectories. ... we identified the time lag that minimized this error. Finally, we calculated the RMSE and Pearson correlation between the adjusted kinematic trajectories from OMC and MMC for each trial to assess the agreement between the two systems."

    The headline agreement metrics are computed after fitting and removing a per-trial constant offset and after choosing the time lag that minimizes RMSE on the same trial. For shoulder flexion the removed bias is -23.55° (IQR -26.57 to -20.40), while Table I reports post-removal RMSE of only 2.39°-2.53°. The abstract's 'RMSE values ranging from 2-5 degrees' and the conclusion that MMC is 'approaching the accuracy of marker-based methods' therefore rest on residuals that are, by construction, insensitive to systematic offsets up to tens of degrees. The bias is honestly reported in Table I, so this is not a hidden fit, but the central accuracy claim is a post-fit waveform-similarity statement rather than a standalone absolute-error prediction.

full rationale

The paper's core comparison is benchmarked against an external system (OMC), not against its own equations or a self-citation chain. The end-to-end MMC algorithm is adopted from prior work by co-authors [16], [17], but the present study's contribution is the OMC-versus-MMC comparison; the OMC data are independent of the MMC optimization, so the central finding is not forced by the method's own inputs. However, the quantitative accuracy claim deserves one caveat that borders on circularity. As quoted above, all RMSE and correlation values in Table I are computed after subtracting the per-trial mean difference and after time-shifting to minimize RMSE. This makes the reported 2-5° joint-angle RMSE a measure of waveform agreement after removing systematic offsets, not a measure of absolute joint-angle error; for shoulder flexion the removed offset averages -23.55°. The paper discloses the bias and discusses its likely sources, so the fitted parameters are not hidden, but the abstract's phrasing 'approaching the accuracy of marker-based methods' presents the post-fit residual as standalone accuracy. A second, explicitly disclosed limitation is that drinking-task phase classification uses OMC data for both systems; phase-dependent quality measures therefore are not fully independent MMC estimates, and the authors correctly state that 'MMC phase classification still requires validation for standalone application.' This is a limitation rather than a circular derivation because the paper does not use it to claim standalone MMC phase classification. No uniqueness theorem, ansatz-by-citation, or renamed known result is present. On balance, the derivation is externally anchored and reproducible, so the circularity score is low.

Assumptions & free parameters 3 free parameters · 4 assumptions · 0 invented entities

The study introduces no new physical entities; its load-bearing assumptions are about measurement reference validity, model equivalence, and the legitimacy of bias and lag corrections. The per-trial bias and time lag are the most consequential fitted quantities because they are removed before the headline agreement scores are computed.

free parameters (3)
  • Per-trial static bias offset = Median shoulder flexion bias -23.55 degrees, elbow extension -7.52 degrees (per trial, varies)
    Computed as the difference in means between MMC and OMC per trajectory and removed before computing RMSE and correlation; the reported agreement is conditional on this correction.
  • Per-trial time lag = Typically 0.00 to 0.03 seconds, optimized within plus or minus 0.25 seconds
    Chosen by shifting one trajectory to minimize RMSE before computing r and RMSE; this can inflate agreement.
  • MMC body scale, marker offset, and joint-angle MLP parameters = Not reported individually
    Learned end-to-end from video via bilevel optimization; they define the MMC kinematics being compared. They are fitted to video, not to OMC, so they do not directly force agreement.
assumptions (4)
  • domain assumption OMC is a valid reference standard for human movement kinematics
    The paper treats OMC as the gold standard while acknowledging it is not absolute ground truth (Discussion, Limitations). If OMC marker errors are large, the agreement metrics do not establish MMC accuracy.
  • domain assumption The OpenSim and MuJoCo biomechanical models are sufficiently equivalent for the relevant degrees of freedom
    The authors state no major differences in relevant DOF between models, but model differences are a leading explanation for the systematic angle offsets (Section II.C.2 and Discussion).
  • ad hoc to paper Static bias and time-lag alignment are valid nuisance corrections
    The central agreement metrics are computed after subtracting per-trial mean offsets and time-shifting to minimize RMSE (Section II.D.1). The validity of treating large offsets as nuisance is assumed, not tested.
  • domain assumption Phase classification from OMC data transfers correctly to MMC trajectories
    Movement quality measures use phase boundaries computed from OMC wrist and cup markers (Section II.D.2), so MMC's standalone phase classification remains unvalidated.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Differentiable Biomechanics for Markerless Motion Capture in Upper Limb Stroke Rehabilitation: A Comparison with Optical Motion Capture." pith.science (2026). https://pith.science/paper/GZKOKHFR

@misc{pith2026241114992,
  author       = {Pith},
  title        = {Pith review of: Differentiable Biomechanics for Markerless Motion Capture in Upper Limb Stroke Rehabilitation: A Comparison with Optical Motion Capture},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/GZKOKHFR}},
  note         = {Machine review of arXiv:2411.14992}
}
read the original abstract

Marker-based Optical Motion Capture (OMC) paired with biomechanical modeling is currently considered the most precise and accurate method for measuring human movement kinematics. However, combining differentiable biomechanical modeling with Markerless Motion Capture (MMC) offers a promising approach to motion capture in clinical settings, requiring only minimal equipment, such as synchronized webcams, and minimal effort for data collection. This study compares key kinematic outcomes from biomechanically modeled MMC and OMC data in 15 stroke patients performing the drinking task, a functional task recommended for assessing upper limb movement quality. We observed a high level of agreement in kinematic trajectories between MMC and OMC, as indicated by high correlations (median r above 0.95 for the majority of kinematic trajectories) and median RMSE values ranging from 2-5 degrees for joint angles, 0.04 m/s for end-effector velocity, and 6 mm for trunk displacement. Trial-to-trial biases between OMC and MMC were consistent within participant sessions, with interquartile ranges of bias around 1-3 degrees for joint angles, 0.01 m/s in end-effector velocity, and approximately 3mm for trunk displacement. Our findings indicate that our MMC for arm tracking is approaching the accuracy of marker-based methods, supporting its potential for use in clinical settings. MMC could provide valuable insights into movement rehabilitation in stroke patients, potentially enhancing the effectiveness of rehabilitation strategies.

Figures

Figures reproduced from arXiv: 2411.14992 by the authors.

Figure 1
Figure 1. Camera perspectives of the five webcams used for [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. Visualization of biomechnical modelling. left: OMC [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗
Figure 3
Figure 3. Exemplary kinematic trajectories of participant p01, [PITH_FULL_IMAGE:figures/full_fig_p004_3.png] view at source ↗
Figures from the paper (1 more)
Figure 4
Figure 4. Figure 4: Correlation plot for each measure, comparing two-stage OMC and end-to-end MMC. Each data point represents a [PITH_FULL_IMAGE:figures/full_fig_p006_4.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

34 extracted references · 33 canonical work pages

  1. [13]

    Conclusion or Illusion: Quantifying Un- certainty in Inverse Analyses From Marker-Based Motion Capture due to Errors in Marker Registration and Model Scaling,

    T. K. Uchida and A. Seth, “Conclusion or Illusion: Quantifying Un- certainty in Inverse Analyses From Marker-Based Motion Capture due to Errors in Marker Registration and Model Scaling,” Frontiers in Bioengineering and Biotechnology , vol. 10, May 2022. Publisher: Frontiers

    work page 2022

  2. [1]

    Standardized measurement of sensori- motor recovery in stroke trials: Consensus-based core recommendations from the Stroke Recovery and Rehabilitation Roundtable,

    G. Kwakkel, N. A. Lannin, K. Borschmann, C. English, M. Ali, L. Churilov, G. Saposnik, C. Winstein, E. E. Van Wegen, S. L. Wolf, J. W. Krakauer, and J. Bernhardt, “Standardized measurement of sensori- motor recovery in stroke trials: Consensus-based core recommendations from the Stroke Recovery and Rehabilitation Roundtable,” International Journal of Stro...

    work page 2017

  3. [2]

    Estimates of the Prevalence of Acute Stroke Impairments and Disability in a Multiethnic Population,

    E. S. Lawrence, C. Coshall, R. Dundas, J. Stewart, A. G. Rudd, R. Howard, . Charles, and D. A. Wolfe, “Estimates of the Prevalence of Acute Stroke Impairments and Disability in a Multiethnic Population,” pp. 1279–1284, 2001

    work page 2001

  4. [3]

    Standardized measurement of quality of upper limb movement af- ter stroke: Consensus-based core recommendations from the Second Stroke Recovery and Rehabilitation Roundtable,

    G. Kwakkel, E. E. H. V . Wegen, J. H. Burridge, C. J. Winstein, L. E. H. v. Dokkum, M. A. Murphy, M. F. Levin, and J. W. Krakauer, “Standardized measurement of quality of upper limb movement af- ter stroke: Consensus-based core recommendations from the Second Stroke Recovery and Rehabilitation Roundtable,” International Journal of Stroke, vol. 14, pp. 783...

    work page 2019

  5. [4]

    Enhancing Brain Plasticity to Promote Stroke Recovery,

    F. Su and W. Xu, “Enhancing Brain Plasticity to Promote Stroke Recovery,” Frontiers in Neurology, vol. 11, p. 554089, Oct. 2020

    work page 2020

  6. [5]

    Neu- roplastic Changes Following Brain Ischemia and their Contribution to Stroke Recovery: Novel Approaches in Neurorehabilitation,

    C. Alia, C. Spalletti, S. Lai, A. Panarese, G. Lamola, F. Bertolucci, F. Vallone, A. Di Garbo, C. Chisari, S. Micera, and M. Caleo, “Neu- roplastic Changes Following Brain Ischemia and their Contribution to Stroke Recovery: Novel Approaches in Neurorehabilitation,” Frontiers in Cellular Neuroscience , vol. 11, Mar. 2017

    work page 2017

  7. [6]

    Movement Quality: A Novel Biomarker Based on Principles of Neuroscience,

    S. Solnik, M. P. Furmanek, and D. Piscitelli, “Movement Quality: A Novel Biomarker Based on Principles of Neuroscience,” Neurorehabil- itation and Neural Repair , vol. 34, pp. 1067–1077, Dec. 2020

    work page 2020

  8. [7]

    What do motor

    M. F. Levin, J. A. Kleim, and S. L. Wolf, “What do motor ”recovery” and ”compensationg” mean in patients following stroke?,” Neuroreha- bilitation and Neural Repair , vol. 23, pp. 313–319, May 2009

    work page 2009

Show all 34 references
  1. [8]

    Kinematic analysis using 3D motion capture of drinking task in people with and without upper-extremity impairments,

    M. A. Murphy, S. Murphy, H. C. Persson, U. B. Bergstr ¨om, and K. S. Sunnerhagen, “Kinematic analysis using 3D motion capture of drinking task in people with and without upper-extremity impairments,” Journal of Visualized Experiments , vol. 2018, Mar. 2018. Publisher: Journal ...

    2018

  2. [9]

    How many trials are needed in kinematic analysis of reach-to-grasp?—A study of the drinking task in persons with stroke and non-disabled controls,

    G. E. Frykberg, H. Grip, and M. A. Murphy, “How many trials are needed in kinematic analysis of reach-to-grasp?—A study of the drinking task in persons with stroke and non-disabled controls,” Journal of NeuroEngineering and Rehabilitation, vol. 18, pp. number–101, Dec

  3. [10]

    Movement kine- matics during a drinking task are associated with the activity capacity level after stroke,

    M. A. Murphy, C. Will ´en, and K. S. Sunnerhagen, “Movement kine- matics during a drinking task are associated with the activity capacity level after stroke,” Neurorehabilitation and Neural Repair, vol. 26, no. 9, pp. 1106–1115, 2012

    2012

  4. [11]

    Moving fluoroscopy-based analysis of THA kinematics during unrestricted activities of daily living,

    F. D’Isidoro, C. Brockmann, B. Friesenbichler, T. Zumbrunn, M. Leu- nig, and S. J. Ferguson, “Moving fluoroscopy-based analysis of THA kinematics during unrestricted activities of daily living,” Frontiers in Bioengineering and Biotechnology , vol. 11, p. 1095845, Apr. 2023

    2023

  5. [12]

    A systematic review of the applications of markerless motion capture (MMC) technology for clinical measurement in rehabilitation,

    W. W. T. Lam, Y . M. Tang, and K. N. K. Fong, “A systematic review of the applications of markerless motion capture (MMC) technology for clinical measurement in rehabilitation,” Dec. 2023. ISSN: 17430003 Is- sue: 1 Publication Title: Journal of NeuroEngineering and Rehabilitat...

    2023

  6. [14]

    Soft tissue artifact causes significant errors in the calculation of joint angles and range of motion at the hip,

    N. M. Fiorentino, P. R. Atkins, M. J. Kutschke, J. M. Goebel, K. B. Foreman, and A. E. Anderson, “Soft tissue artifact causes significant errors in the calculation of joint angles and range of motion at the hip,” Gait & posture , vol. 55, pp. 184–190, June 2017

    2017

  7. [15]

    A Deep Learning Model for Markerless Pose Estimation Based on Keypoint Augmentation: What Factors Influence Errors in Biomechanical Applications?,

    A. V . Ruescas-Nicolau, E. Medina-Ripoll, H. de Rosario, J. Sanchiz Navarro, E. Parrilla, and M. C. Juan Lizandra, “A Deep Learning Model for Markerless Pose Estimation Based on Keypoint Augmentation: What Factors Influence Errors in Biomechanical Applications?,” Sensors, vol....

    1923

  8. [16]

    Optimizing Trajectories and Inverse Kinematics for Biomechanical Analysis of Markerless Motion Capture Data,

    R. J. Cotton, A. DeLillo, A. Cimorelli, K. Shah, J. Peiffer, S. Anarwala, K. Abdou, and T. Karakostas, “Optimizing Trajectories and Inverse Kinematics for Biomechanical Analysis of Markerless Motion Capture Data,” in 2023 International Conference on Rehabilitation Robotics (IC...

    2023

  9. [17]

    Differentiable Biomechanics Unlocks Opportunities for Markerless Motion Capture,

    R. J. Cotton, “Differentiable Biomechanics Unlocks Opportunities for Markerless Motion Capture,” Feb. 2024. arXiv:2402.17192 [cs]

    2024 arXiv

  10. [18]

    Biomechanical Arm and Hand Tracking with Multiview Markerless Motion Capture,

    P. Firouzabadi, W. Murray, A. R. Sobinov, J. Peiffer, K. Shah, L. E. Miller, and R. J. Cotton, “Biomechanical Arm and Hand Tracking with Multiview Markerless Motion Capture,” in 2024 10th IEEE RAS/EMBS International Conference for Biomedical Robotics and Biomechatronics (BioRo...

    2024

  11. [19]

    Full-Body Musculoskeletal Model for Muscle- Driven Simulation of Human Gait,

    A. Rajagopal, C. L. Dembia, M. S. DeMers, D. D. Delp, J. L. Hicks, and S. L. Delp, “Full-Body Musculoskeletal Model for Muscle- Driven Simulation of Human Gait,” IEEE Transactions on Biomedical Engineering, vol. 63, pp. 2068–2079, Oct. 2016

    2016

  12. [20]

    AniposeLib,

    L. Karashchuk, “AniposeLib,” Nov. 2024

    2024

  13. [21]

    PosePipe: Open-Source Human Pose Estimation Pipeline for Rehabilitation Research,

    R. J. Cotton, “PosePipe: Open-Source Human Pose Estimation Pipeline for Rehabilitation Research,” Archives of Physical Medicine and Reha- bilitation, vol. 103, pp. e161–e162, Dec. 2022. Publisher: Elsevier

    2022

  14. [22]

    Learning 3D Human Pose Esti- mation from Dozens of Datasets using a Geometry-Aware Autoencoder to Bridge Between Skeleton Formats,

    I. S ´ar´andi, A. Hermans, and B. Leibe, “Learning 3D Human Pose Esti- mation from Dozens of Datasets using a Geometry-Aware Autoencoder to Bridge Between Skeleton Formats,” 2022. Version Number: 1

    2022

  15. [23]

    EasyMoCap - Make human motion capture easier.,

    “EasyMoCap - Make human motion capture easier.,” 2021

    2021

  16. [24]

    MuJoCo: A physics engine for model-based control,

    E. Todorov, T. Erez, and Y . Tassa, “MuJoCo: A physics engine for model-based control,” in 2012 IEEE/RSJ International Conference on Intelligent Robots and Systems, pp. 5026–5033, Oct. 2012. ISSN: 2153- 0866

    2012

  17. [25]

    MyoSuite: A Contact-rich Simulation Suite for Musculoskeletal Motor Control,

    V . Caggiano, H. Wang, G. Durandau, M. Sartori, and V . Kumar, “MyoSuite: A Contact-rich Simulation Suite for Musculoskeletal Motor Control,” in Proceedings of The 4th Annual Learning for Dynamics and Control Conference, pp. 492–507, PMLR, May 2022. ISSN: 2640-3498

    2022

  18. [26]

    Rapid bilevel optimization to concurrently solve muscu- loskeletal scaling, marker registration, and inverse kinematic problems for human motion reconstruction,

    K. Werling, M. Raitor, J. Stingel, J. L. Hicks, S. Collins, S. L. Delp, and C. K. Liu, “Rapid bilevel optimization to concurrently solve muscu- loskeletal scaling, marker registration, and inverse kinematic problems for human motion reconstruction,” Aug. 2022

    2022

  19. [27]

    Muscle contributions to propul- sion and support during running,

    S. R. Hamner, A. Seth, and S. L. Delp, “Muscle contributions to propul- sion and support during running,” Journal of Biomechanics , vol. 43, pp. 2709–2716, Oct. 2010

    2010

  20. [28]

    LocoMuJoCo: A Comprehensive Imitation Learning Benchmark for Locomotion,

    F. Al-Hafez, G. Zhao, J. Peters, and D. Tateo, “LocoMuJoCo: A Comprehensive Imitation Learning Benchmark for Locomotion,” 2023. Version Number: 2

    2023

  21. [29]

    Accuracy of a markerless motion capture system in estimating upper extremity kinematics during boxing,

    B. K. Lahkar, A. Muller, R. Dumas, L. Reveret, and T. Robert, “Accuracy of a markerless motion capture system in estimating upper extremity kinematics during boxing,” Frontiers in Sports and Active Living, vol. 4, p. 939980, July 2022

    2022

  22. [30]

    Validation of upper extremity kinematics using Markerless motion capture,

    R. M. Hansen, S. L. Arena, and R. M. Queen, “Validation of upper extremity kinematics using Markerless motion capture,” Biomedical Engineering Advances, vol. 7, p. 100128, June 2024

    2024

  23. [31]

    Computer vision for kinematic metrics of the drinking task in a pilot study of neurotypical participants,

    J. Huber, S. Slone, and J. Bae, “Computer vision for kinematic metrics of the drinking task in a pilot study of neurotypical participants,” Scientific Reports, vol. 14, p. 20668, Sept. 2024. Publisher: Nature Publishing Group

    2024

  24. [32]

    Inter-session repeatability of markerless motion capture gait kinematics,

    R. M. Kanko, E. Laende, W. S. Selbie, and K. J. Deluzio, “Inter-session repeatability of markerless motion capture gait kinematics,” Journal of Biomechanics, vol. 121, p. 110422, May 2021

    2021

  25. [33]

    Upper limb movement quality measures: comparing IMUs and optical motion capture in stroke patients performing a drinking task,

    T. Unger, R. d. S. Ribeiro, M. Mokni, T. Weikert, J. Pohl, A. Schwarz, J. P. O. Held, L. Sauerzopf, B. K ¨uhnis, E. Gavagnin, A. R. Luft, R. Gassert, O. Lambercy, C. A. Easthope, and J. G. Sch ¨onhammer, “Upper limb movement quality measures: comparing IMUs and optical motion ...

    2024

  26. [2021]

    Publisher: BioMed Central Ltd

Pith tools

Reviewed August 12, 2026 · model on record in the stance chip above.