Pith. sign in

REVIEW 4 major objections 6 minor 63 references

TrackStudio: An Integrated Toolkit for Markerless Tracking

T0 review · 4 major / 6 minor · reviewed 2026-08-03 · deepseek-v4-flash

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

Pith's one-line read The paper argues that a no-code, GUI-based toolkit can make markerless 3D tracking practical for non-experts, achieving stable performance across varied real-world settings.

desk verdict Useful engineering contribution with a seriously overstated accuracy claim: the reported 'triangulation error' is a self-reprojection residual, not a ground-truth error. read the letter →

arxiv 2511.07624 v3 pith:L2EHNLBA submitted 2025-11-10 cs.CV q-bio.QM

Hristo Dimitrov , Viktorija Pavalkyte , Giulia Dominijanni , Tamar R. Makin This is my paper
classification cs.CVq-bio.QM
keywords markerlessmotiontrackingposeestimation3DtriangulationcameracalibrationGUItoolkitno-codehandaccessibility
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 argues that the main barrier to wider use of markerless motion tracking is procedural, not algorithmic: existing tools are accurate but require coding, installation, and calibration expertise. TrackStudio is proposed as a solution—a GUI-based pipeline that combines established open-source components for 2D pose estimation, multi-camera calibration, 3D triangulation, and visualization into one installable application. The authors validate it on 76 participants across three environments (seated, supine, mixed) using webcams or high-resolution cameras, reporting average inter-frame correlations above 0.98 and median triangulation errors below 13.6 mm for hand tracking. If the central claim holds, non-experts could collect usable 3D kinematic data in labs, classrooms, or field settings without specialized training.

What carries the argument

The carrying mechanism is the pre-packaged GUI workflow itself: it chains a 2D landmark-detection model (which supplies per-camera 'virtual markers'), a checkerboard calibration step that estimates each camera's position and lens distortion, and a triangulation step that reconstructs 3D coordinates from corresponding 2D detections across synchronized cameras. The GUI enforces the correct order and hides the underlying code, while bundled virtual environments prevent the dependency and version conflicts that typically block non-experts. The validation's internal quality metrics—inter-frame correlation, movement smoothness, and 2D reprojection error converted to millimeters—are used to demonst

What would settle it

Record a scene where true 3D positions are known—such as a wand with two markers at a fixed distance, or a participant wearing a marker-based suit—run TrackStudio on the same footage, and compare its 3D landmark coordinates to the measured truth; if median absolute error is large despite low reprojection error, the reliability claim is falsified.

Watch

Extended reading notes

Core claim

TrackStudio's central discovery is that integrating off-the-shelf tracking components behind a GUI, with pinned software environments and a practical user manual, produces a pipeline that delivers stable tracking in imperfect real-world conditions. Across all validation setups, inter-frame correlations exceeded 0.98, movement smoothness followed expected values for simple versus complex actions, and median 3D reprojection errors were low: 13.6 mm for seated hand tracking with webcams, 8.95 mm supine, and 4.41 mm for hands in the mixed high-resolution setup, with errors for face (9.52 mm) and arms (17.6 mm) higher but within the authors' cited state-of-the-art range. The toolkit is modular, s

Load-bearing premise

The premise that the pipeline's own internal metrics—reprojection error and frame-to-frame correlation—stand in for true tracking accuracy; if systematic 3D bias or drift keeps these internal errors low while the true positions are off, the 'reliable performance' claim does not follow.

Editorial extensions

If this is right

  • Behavioral, clinical, and biomechanical researchers can run 3D tracking with commodity webcams and no programming staff, lowering the cost barrier for small labs.
  • Educational and pilot settings, where setup effort often outweighs expected gains, become practical places to collect motion data.
  • The modular design lets future users swap in better 2D/3D components as they appear, without rebuilding the workflow.
  • The same pipeline works beyond hands—face and upper-body landmarks were tracked—so a single toolkit can serve multiple measurement needs.
  • Pinned software environments, while limiting updates, offer a reproducibility advantage for longitudinal studies.

Reading between the lines

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

  • Because the validation measures reprojection error rather than absolute error against known 3D positions, the paper's accuracy claim is best read as 'internally stable'; a ground-truth benchmark would be needed before relying on the toolkit for clinical-grade kinematics.
  • The requirement that a body part stay visible to at least two (ideally three) cameras means setups with severe or unpredictable occlusion—full-body motion, crowded scenes—will still need careful camera placement or more cameras, a constraint the authors acknowledge.
  • The toolkit's strategy of removing 'procedural friction' could be applied more broadly: packaging expert CV tools behind GUIs with bundled environments may boost adoption more than incremental accuracy improvements.
  • A testable next step is to run TrackStudio and a marker-based system simultaneously on the same tasks; a close match on joint angles and distances would extend the claim from usability to measurement validity.
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. TrackStudio is presented as an integrated, GUI-based toolkit for markerless 3D motion tracking that combines MediaPipe 2D detection, Anipose-inspired calibration and triangulation, and video-processing utilities into a package that requires no programming. The paper's validation is based on data from 76 (or 75) participants across three recording environments: seated webcam, supine webcam, and mixed high-resolution camera setups. The headline quantitative claims are average inter-frame correlations above 0.98 and 'triangulation errors' below 13.6 mm for hand tracking, plus smoothness (LDJ) values in expected ranges. The authors also provide a user manual covering installation, camera synchronization, calibration, and practical pitfalls. The central claim is that TrackStudio offers reliable performance for non-experts in varied settings.

Significance. If the reported validation were valid, the paper would fill a genuine accessibility gap: a preconfigured, GUI-based pipeline with packaged environments, documentation, and modular design is a useful contribution for behavioral and clinical researchers without programming expertise. The engineering effort is real and the manual is unusually practical. The code is made available on GitHub and packaged environments on FigShare, which supports reproducibility of the toolkit itself. However, the quantitative validation in the current form does not support the accuracy claims in the abstract: the 'triangulation error' is a self-reprojection residual, inter-frame correlation measures smoothness rather than positional accuracy, and no ground-truth comparison is reported. These issues are load-bearing because the abstract and conclusion rely on 'reliable performance' and specific error numbers. The contribution is defensible as an accessibility/usability engineering paper, but the accuracy claims must be either substantiated with independent ground truth or substantially reframed.

major comments (4)
  1. [Methods, 'Pipeline Performance Metrics'; Results; Abstract] The '3D error approximation' is defined as 'the error between the marker predicted by the 2D annotation and the 3D marker reprojected back into 2D space' and then converted to mm using camera distance and lens parameters. This is a 2D reprojection residual, not a 3D triangulation error. A systematic bias in calibration or detection can shift the reconstructed 3D point while its 2D reprojection remains close to the detected 2D point, so small residuals do not imply accurate 3D positions. The abstract's statement 'average triangulation errors remained low (<13.6mm for hand tracking)' and the Results' comparison with per-joint 3D errors from refs 33–34 are therefore not supported by the reported metric.
  2. [Methods, 'Pipeline Performance Metrics'; Results, Figures 3–5] The inter-frame Pearson correlation is computed on the marker's 3D vector magnitude over 5 ms intervals and explicitly discards direction changes. This metric measures local smoothness and continuity, not positional accuracy. A trajectory with a constant offset or a slow, systematic drift can still produce r > 0.98. Consequently, 'stable and consistent tracking' (abstract) conflates smoothness with accuracy. The paper needs an independent positional accuracy metric, such as comparison to a marker-based system or manually annotated ground-truth points, before claiming 'reliable performance.'
  3. [Methods, 'Pipeline Performance Metrics'; Discussion] The paper explicitly states 'for more details on ground truth comparison, see Tony Hii et al., 2023' but reports no ground-truth comparison for TrackStudio data. The cited work is a different study. Without any external reference, the validation is self-referential: the pipeline is evaluated against its own 2D detections and its own 3D reconstruction. This may demonstrate internal consistency but cannot establish absolute tracking accuracy. The authors should either add a ground-truth validation experiment (e.g., tracking known 3D points or comparing to a marker-based system in a subset of recordings) or revise the abstract and conclusions to avoid claiming accuracy beyond internal consistency.
  4. [Experimental Validation; Table 1; Abstract] There is an inconsistency in participant counts and demographics. The abstract states 76 participants, but the text reports 45 (seated) + 25 (supine) + 5 (mixed) = 75. Table 1 lists 'Suppine' as 26 participants (14 female, mean age 54.7 ± 12.8), while the Methods text says 25 participants (12 female, mean age 42.7 ± 3.91). This discrepancy affects the total participant count and the reported supine statistics. Please correct and ensure all derived numbers are consistent.
minor comments (6)
  1. [Abstract] Typo: 'lightning' should be 'lighting' in the phrase 'challenging conditions for body position, lightning, and space and obstructions.'
  2. [Methods, 'Pipeline Performance Metrics'] The metric is called 'inter-frame Person's cross-correlation'; it should be 'Pearson's cross-correlation.'
  3. [Table 1] Header 'Suppine' is a typo for 'Supine.'
  4. [Introduction, Toolkit Overview] The text mentions 'issues due to operating systems and silicone'; likely 'silicon' was intended.
  5. [Appendix, Supplementary Figure 5] The caption refers to 'TrakStudio's Video Trimming window'; should be 'TrackStudio.'
  6. [Results, Mixed setup] The statement 'no significant day-to-day or device-wearing differences were detected (0.98 > p > 0.14, 0.09< BF10 < 0.95)' relies on a very small sample (n=5) and Bayes factors near 1; this should be presented as weak evidence, not as confirmation of stability.

Circularity Check

1 steps flagged · score 6.0 of 10

The 'triangulation error' used to validate TrackStudio is defined as the pipeline's own 2D reprojection residual, so the headline accuracy claim reduces to an internal-consistency check rather than an independent measure of 3D error.

  1. self definitional [Methods, 'Pipeline Performance Metrics'; also Abstract and Results]
    "the 3D error approximation is representative of the amount of error incurred during the 3D triangulation. It is computed by calculating the error between the marker predicted by the 2D annotation and the 3D marker reprojected back into 2D space."

    The '3D error' is defined as the distance between the 2D annotation (the input to triangulation) and the reprojection of the 3D point reconstructed from that same annotation. This is a reprojection residual of the pipeline's own reconstruction, not a ground-truth-referenced triangulation error. The abstract's claim that 'average triangulation errors remained low (<13.6mm)' is therefore a statement about internal consistency between the pipeline's 2D detections and its own 3D output, by construction. A systematically biased 3D reconstruction can still reproject close to the detected 2D points, so the metric can remain small even when the 3D position is wrong. No independent ground-truth comparison is reported for TrackStudio data, making the central accuracy claim partially self-referential

full rationale

The main circular step is the validation metric. TrackStudio's '3D error approximation' is explicitly computed by reprojecting the pipeline's own 3D triangulation back into 2D and comparing it with the 2D detections from which that triangulation was derived. This is not a true 3D positional error; it is a residual that measures how well the 3D reconstruction explains its own 2D inputs. The abstract's quantitative 'triangulation error' claim therefore reduces, by definition, to an internal-consistency measure. The paper also compares these residuals to external benchmarks that report ground-truth per-joint position error, but that comparison does not rescue the circularity because it equates incompatible quantities. The engineering and accessibility contributions are credible, and the underlying MediaPipe/Anipose components have external validation, so the paper is not wholly circular; however, the central numerical validation claim is self-referential. No parameter fitting or self-citation chain is involved, which keeps the score at 6 rather than higher.

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

The central claims rest on trust in the underlying open-source trackers and on treating self-reprojection error as accuracy. No free parameters are fitted to the validation outcomes; thresholds such as 10 mm/30 mm error bins are analytic choices, and standard camera calibration parameters are not ad hoc free parameters.

assumptions (4)
  • domain assumption MediaPipe provides accurate 2D landmark localization in the tested conditions.
    All 2D and downstream 3D results inherit MediaPipe's detections; the toolkit does not validate or correct them.
  • domain assumption Anipose's camera calibration and triangulation produce correct 3D geometry.
    3D reconstruction and the reported reprojection errors depend on Anipose's calibration and triangulation routines.
  • ad hoc to paper Self-reprojection error is a valid proxy for true tracking accuracy.
    The Methods define 3D error as the residual between 2D detections and reprojected 3D markers; this only measures triangulation self-consistency, not absolute accuracy.
  • domain assumption Camera frame synchronization via LED cues or software timestamps is accurate enough for 3D reconstruction.
    Triangulation assumes corresponding frames; synchronization accuracy is recommended via LED but not quantitatively validated in the paper.

how reviews work

0 comments
Cite this review

Pith. "Pith review of TrackStudio: An Integrated Toolkit for Markerless Tracking." pith.science (2026). https://pith.science/paper/L2EHNLBA

@misc{pith2026251107624,
  author       = {Pith},
  title        = {Pith review of: TrackStudio: An Integrated Toolkit for Markerless Tracking},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/L2EHNLBA}},
  note         = {Machine review of arXiv:2511.07624}
}
read the original abstract

Markerless motion tracking has advanced rapidly in the past 10 years and currently offers powerful opportunities for behavioural, clinical, and biomechanical research. While several specialised toolkits provide high performance for specific tasks, using existing tools still requires substantial technical expertise. There remains a gap in accessible, integrated solutions that deliver sufficient tracking for non-experts across diverse settings. TrackStudio was developed to address this gap by combining established open-source tools into a single, modular, GUI-based pipeline that works out of the box. It provides video recording preprocessing, recording synchronisation, automatic 2D and 3D pose estimation, and visualisation without requiring any programming skills. We supply a user guide with practical advice for video acquisition, camera calibration, video synchronisation, and experimental setup, alongside documentation of common pitfalls and how to avoid them. To validate the toolkit, we tested its performance across three environments using either low-cost webcams or high-resolution cameras, including challenging conditions for body position, lighting, space, and obstructions. Across 76 participants, average inter-frame correlations exceeded 0.98 and average triangulation errors remained low (<13.6mm for hand tracking), demonstrating stable and consistent tracking. We further show that the same pipeline can be extended beyond hand tracking to other body and face regions. TrackStudio provides a practical, accessible route into markerless tracking for researchers or laypeople who need reliable performance without specialist expertise.

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

63 extracted references · 1 canonical work pages

  1. [1]

    Winter, D. A. Biomechanics and Motor Control of Human Movement. (John Wiley & Sons, 2009)

    2009

  2. [2]

    Roemmich, R. T. et al. Interlimb coordination is impaired during walking in persons with Parkinson’s disease. Clinical Biomechanics 28, 93–97 (2013)

    2013

  3. [5]

    & Reijne, M

    van der Kruk, E. & Reijne, M. M. Accuracy of human motion capture systems for sport applications; state-of-the-art review. Eur J Sport Sci 18, 806–819 (2018)

    2018

  4. [6]

    Filippeschi, A. et al. Survey of Motion Tracking Methods Based on Inertial Sensors: A Focus on Upper Limb Human Motion. Sensors 17, (2017)

    2017

  5. [7]

    & García-Domínguez, J

    García-de-Villa, S., Casillas-Pérez, D., Jiménez-Martín, A. & García-Domínguez, J. J. Inertial Sensors for Human Motion Analysis: A Comprehensive Review. IEEE Trans Instrum Meas 72, 1–39 (2023)

    2023

  6. [8]

    Franz, A. M. et al. Electromagnetic Tracking in Medicine—A Review of Technology, Validation, and Applications. IEEE Trans Med Imaging 33, 1702–1725 (2014)

    2014

  7. [9]

    & Mathis, M

    Mathis, A., Schneider, S., Lauer, J. & Mathis, M. W. A Primer on Motion Capture with Deep Learning: Principles, Pitfalls, and Perspectives. Neuron 108, 44–65 (2020)

    2020

  8. [10]

    M., Laende, E

    Kanko, R. M., Laende, E. K., Davis, E. M., Selbie, W. S. & Deluzio, K. J. Concurrent assessment of gait kinematics using marker-based and markerless motion capture. J Biomech 127, 110665 (2021)

    2021

Show all 63 references
  1. [11]

    Scataglini, S. et al. Accuracy, Validity, and Reliability of Markerless Camera- Based 3D Motion Capture Systems versus Marker-Based 3D Motion Capture Systems in Gait Analysis: A Systematic Review and Meta-Analysis. Sensors 24, (2024)

    2024

  2. [12]

    Wang, J. et al. Deep High-Resolution Representation Learning for Visual Recognition. IEEE Trans Pattern Anal Mach Intell 43, 3349–3364 (2021)

    2021

  3. [13]

    & Sheikh, Y

    Cao, Z., Hidalgo, G., Simon, T., Wei, S.-E. & Sheikh, Y. OpenPose: Realtime Multi- Person 2D Pose Estimation Using Part Affinity Fields. IEEE Trans Pattern Anal Mach Intell 43, 172–186 (2021)

    2021

  4. [14]

    Zhang, F. et al. MediaPipe Hands: On-device Real-time Hand Tracking. CoRR abs/2006.10214, (2020)

    2006 arXiv

  5. [15]

    Karashchuk, P. et al. Anipose: A toolkit for robust markerless 3D pose estimation. Cell Rep 36, 109730 (2021)

    2021

  6. [16]

    & Scherberger, H

    Sheshadri, S., Dann, B., Hueser, T. & Scherberger, H. 3D reconstruction toolbox for behavior tracked with multiple cameras. J Open Source Softw 5, 1849 (2020)

    2020

  7. [17]

    Mathis, A. et al. DeepLabCut: markerless pose estimation of user-defined body parts with deep learning. Nat Neurosci 21, 1281–1289 (2018)

    2018

  8. [18]

    Real-time AI motion capture in your browser | RADiCAL Live

    RADiCAL Motion. Real-time AI motion capture in your browser | RADiCAL Live. https://radicalmotion.com/live (2025)

    2025

  9. [19]

    M., Costantino, M., Freud, E

    Mulla, D. M., Costantino, M., Freud, E. & Michaels, J. A. ATHENA: Automatically Tracking Hands Expertly with No Annotations. bioRxiv https://doi.org/10.1101/2025.08.12.669753 (2025) doi:10.1101/2025.08.12.669753

    2025 doi

  10. [20]

    Uhlrich, S. D. et al. OpenCap: Human movement dynamics from smartphone videos. PLoS Comput Biol 19, 1–26 (2023)

    2023

  11. [21]

    MoveNet: Ultra Fast and Accurate Pose Detection Model

    TensorFlow. MoveNet: Ultra Fast and Accurate Pose Detection Model. https://www.tensorflow.org/hub/tutorials/movenet (2024)

    2024

  12. [22]

    OpenMMLab Pose Estimation Toolbox and Benchmark

    MMPose Contributors. OpenMMLab Pose Estimation Toolbox and Benchmark. https://github.com/open-mmlab/mmpose (2020)

    2020

  13. [23]

    Bittner, M. et al. Towards Single Camera Human 3D-Kinematics. Sensors 23, (2023)

    2023

  14. [24]

    Tony Hii, C. S. et al. Frontal Plane Gait Analysis using Pose Estimation Models. in 2nd IEEE National Biomedical Engineering Conference, NBEC 2023 1–6 (Institute of Electrical and Electronics Engineers Inc., 2023). doi:10.1109/NBEC58134.2023.10352623

    2023

  15. [25]

    & Dambre, J

    De Coster, M., Rushe, E., Holmes, R., Ventresque, A. & Dambre, J. Towards the extraction of robust sign embeddings for low resource sign language recognition. ArXiv http://arxiv.org/abs/2306.17558 (2023)

    2023 arXiv

  16. [26]

    & Priano, L

    Amprimo, G., Ferraris, C., Masi, G., Pettiti, G. & Priano, L. GMH-D: Combining Google MediaPipe and RGB-Depth Cameras for Hand Motor Skills Remote Assessment. in Proceedings - 2022 IEEE International Conference on Digital Health, ICDH 2022 132–141 (Institute of Electrical and ...

    2022

  17. [27]

    Dowdall, L. et al. Developing a Sensory Representation of an Artificial Body Part. bioRxiv https://doi.org/10.1101/2025.06.16.658246 (2025) doi:10.1101/2025.06.16.658246

    2025 doi

  18. [28]

    & Burdet, E

    Balasubramanian, S., Melendez-Calderon, A. & Burdet, E. A Robust and Sensitive Metric for Quantifying Movement Smoothness. IEEE Trans Biomed Eng 59, 2126– 2136 (2012)

    2012

  19. [29]

    & Hogan, N

    Flash, T. & Hogan, N. The coordination of arm movements: an experimentally confirmed mathematical model. Journal of Neuroscience 5, 1688–1703 (1985)

    1985

  20. [30]

    Bayle, N. et al. Comparison of Various Smoothness Metrics for Upper Limb Movements in Middle-Aged Healthy Subjects. Sensors 23, (2023)

    2023

  21. [31]

    Engdahl, S. M. & Gates, D. H. Reliability of upper limb movement quality metrics during everyday tasks. Gait Posture 71, 253–260 (2019)

    2019

  22. [32]

    & Hermsdörfer, J

    Gulde, P. & Hermsdörfer, J. Smoothness Metrics in Complex Movement Tasks. Front Neurol Volume 9-2018, (2018)

    2018

  23. [33]

    & Pribanić, T

    Bartol, K., Bojanić, D., Petković, T. & Pribanić, T. Generalizable Human Pose Triangulation. Preprint at https://arxiv.org/abs/2110.00280 (2022)

    2022 arXiv

  24. [34]

    & Malkov, Y

    Iskakov, K., Burkov, E., Lempitsky, V. & Malkov, Y. Learnable Triangulation of Human Pose. in International Conference on Computer Vision (ICCV) (2019)

    2019

  25. [35]

    JASP (Version 0.95.4)[Computer software]

    JASP Team. JASP (Version 0.95.4)[Computer software]. Preprint at https://jasp- stats.org/ (2025)

    2025

  26. [36]

    Using Bayes to get the most out of non-significant results

    Dienes, Z. Using Bayes to get the most out of non-significant results. Front Psychol Volume 5-2014, (2014)

    2014

  27. [37]

    Wetzels, R. et al. Statistical Evidence in Experimental Psychology: An Empirical Comparison Using 855 t Tests. Perspectives on Psychological Science 6, 291– 298 (2011)

    2011

  28. [38]

    M., Stefan, A

    van Doorn, J., Aust, F., Haaf, J. M., Stefan, A. M. & Wagenmakers, E.-J. Bayes Factors for Mixed Models: Perspective on Responses. Comput Brain Behav 6, 127–139 (2023)

    2023

  29. [39]

    & Wagenmakers, E.-J

    Keysers, C., Gazzola, V. & Wagenmakers, E.-J. Using Bayes factor hypothesis testing in neuroscience to establish evidence of absence. Nat Neurosci 23, 788– 799 (2020)

    2020

  30. [40]

    Computer software

    Anaconda Software Distribution. Computer software. Vers. 2-2.4.0. Anaconda. https://anaconda.com (2016)

    2016

  31. [41]

    OBS Project Contributors

    Bailey, H. OBS Project Contributors. Open Broadcasting Software https://www.obsproject.org/ (2017)

    2017

  32. [42]

    Kothe, C. et al. The Lab Streaming Layer for Synchronized Multimodal Recording. Imaging Neuroscience 3, IMAG.a.136 (2025). Acknowledgements This work was supported by intramural funding from the Medical Research Council UK (MC_UU_00030/10) to TRM. We would like to thank to Luc...

    2025

  33. [43]

    press any key

    Set up MediaPipe virtual environment → prompts “press any key” to continue

  34. [44]

    press any key

    Set up Anipose virtual environment (this is the longest step) and runs a quick installation test → prompts “press any key.”

  35. [45]

    press any key

    Update TrackStudio files → prompts “press any key.”

  36. [46]

    press any key,

    Finalise the setup → final “press any key,” and you’re done. Note: The first launch of the GUI may take a few extra seconds while initial variables are created. Supplementary Figure 1. Installation window visualisation, indicating where each step takes place. After installatio...

  37. [47]

    Original Recording Directory

    First time using this folder – new Saving Directory (Supplementary Figure 4A) • If you’ve never run TrackStudio in the selected directory, the first processing step you run (e.g., Video Trimming or 2D Annotation) will: o Recreate your original folder structure (e.g., participa...

  38. [48]

    Saving Directory

    Continuing in an existing MLT folder (Supplementary Figure 4B) • This folder already has results from Video Trimming and/or 2D Annotation and later steps. • TrackStudio will use it as your working folder and save any new results alongside the existing ones. • No new directorie...

  39. [49]

    Selecting Saving Directory for TrackStudio

    (Always Necessary on Launch!) Configure – every time you open the GUI, you need to complete a configuration step (see Supplementary Figure 3A and B), consisting of selecting: a. Saving Directory - location where TrackStudio will stored all your results (see “Selecting Saving D...

  40. [50]

    Best practices for LED sync

    Video Trimming – use this if you need to trim your videos and/or synchronise them using an LED(s) light. When you click Video Trimming, choose a mode: a. Manual • Pick a video, set start and end frames (use the arrow controls or type a number), then click Trim. • The trimmed v...

  41. [51]

    Draw a small square around the LED

    A frame appears. Draw a small square around the LED

  42. [52]

    Press Enter to move to the next camera view and repeat

  43. [53]

    ROI Selected

    When all cameras are marked, you will see “ROI Selected” – click Trim. • Trimmed videos are saved under videos-raw/ in your Saving Directory. Advanced options for Automatic (Lights) video trimming • Multiple trims within a video – set how many ON–OFF trials/events are inside o...

  44. [54]

    Change Board Parameters

    Camera Calibration (required for 3D) – the Camera Calibration window, contains 2 sub-windows “Change Board Parameters” and “Calibrate”: a. Change Board Parameters – select only if you used a non-default calibration board (size, type, etc. – see Anipose documentation for detail...

  45. [55]

    • Select a folder that contains your trimmed (and synchronised if 3D required later on) videos

    2D Annotation – this step performs 2D MLT and creates files containing 2D coordinates of virtual markers corresponding to the body part you have chosen in the configuration section. • Select a folder that contains your trimmed (and synchronised if 3D required later on) videos....

  46. [56]

    • Requirements: 2D Annotation and a matching Calibration completed for the selected folder directory

    3D Annotation – combines 2D Annotation results across cameras (using your calibration) to produce 3D coordinates. • Requirements: 2D Annotation and a matching Calibration completed for the selected folder directory. • Select a folder (within your Saving Directory) that shares ...

  47. [57]

    • Select the folder (within your Saving Directory) that contains the 2D Annotation results for the videos you want to review

    2D Video Labelling (optional) – overlays 2D Annotation results onto their respective videos, visualising the results of the tracked body parts. • Select the folder (within your Saving Directory) that contains the 2D Annotation results for the videos you want to review. • Usefu...

  48. [58]

    Video Recordings

    3D Video Labelling (optional) – creates a tiled video showing the original views with overlayed 3D Annotation results for visualisation. • Select the folder (within your Saving Directory) that contains the 3D Annotation results for the videos you want to review. • Note: if a r...

  49. [59]

    Poor lighting can prevent the calibration board from being properly detected

    Ensure the environment is well-lit. Poor lighting can prevent the calibration board from being properly detected

  50. [60]

    Glossy paper, clear tape, or laminating the board may introduce reflections that interfere with detection

    Avoid reflections on the printed calibration board. Glossy paper, clear tape, or laminating the board may introduce reflections that interfere with detection

  51. [61]

    Some printers automatically resize prints to fit the page, which may compromise calibration accuracy

    Verify the calibration board is printed at the correct scale and with high contrast. Some printers automatically resize prints to fit the page, which may compromise calibration accuracy. The dark areas of the board should be fully black to enhance feature detection

  52. [62]

    This prevents hand occlusions

    Leave sufficient margins for holding the board. This prevents hand occlusions. Mounting a handle at the back of the board is recommended

  53. [63]

    This type is more tolerant to occlusions and partial views

    Use a ChArUco calibration board. This type is more tolerant to occlusions and partial views. Not all board corners must be visible in every camera view

  54. [64]

    For example, with three cameras, a total calibration duration of approximately 3 minutes is advised

    We recommend recording calibration videos for at least 1 minute per camera. For example, with three cameras, a total calibration duration of approximately 3 minutes is advised

  55. [65]

    Position the board so that all cameras have an unobstructed view

    Monitor the recording during calibration. Position the board so that all cameras have an unobstructed view. This should be your starting and ending position. a. Begin recording and keep the board visible to all cameras for approximately 20 seconds. b. Then, move the board clos...

Pith tools

Reviewed August 3, 2026 · model on record in the stance chip above.