Pith. sign in

REVIEW 1 major objections 2 minor 20 references

A 3D-Printable Dataset for Fair Testing and Comparisons of Tactile Sensors

T0 review · 1 major / 2 minor · reviewed 2026-06-25 · grok-4.3

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

Pith's one-line read A dataset of six 3D-printable parametric textures supplies the first shared physical benchmark for comparing tactile sensors.

desk verdict The paper supplies a parametric 3D-printable texture dataset and shows its own experiments on printer variance, but the inconsistencies it documents undercut the claim of a reliable cross-printer benchmark. read the letter →

arxiv 2606.25886 v1 pith:T33L7YKW submitted 2026-06-24 cs.RO cs.LG

Dexter R. Shepherd , Nicolas Herzig , Phil Husbands , Andrew Philippides , Chris Johnson , William Kimbell This is my paper · ORCID
classification cs.ROcs.LG
keywords tactilesensors3Dprintabledatasettexturebenchmarkparametricsurfacesreproducibilitysensingsensorcomparisonprinting
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

Existing tactile datasets record readings from one specific sensor on real surfaces, so results cannot be compared fairly across different sensors. The authors generate six surface patterns from combinations of sine-wave and Fourier functions that control spatial frequency, amplitude, and direction. These patterns are printed on three different machines with multiple filaments and then measured with an optical TacTip sensor under fixed contact conditions. Higher-quality prints produce more consistent tactile signatures, allowing neural networks to classify the textures reliably when trained and tested on the same printer. Cross-printer classification fails because geometric differences introduced during fabrication exceed the differences between the intended texture classes.

What carries the argument

Six parametrically generated surface patterns derived from combinations of sine-wave and Fourier-based functions that control spatial frequency, amplitude, and directional structure.

What would settle it

Prints of the same pattern from two different printers that, when scanned by the same sensor under identical conditions, produce signatures differing by more than the typical difference between two distinct sensors would falsify the claim of a usable shared benchmark.

Watch

Extended reading notes

Core claim

The dataset consists of six parametrically generated surface patterns that can be printed reliably enough to serve as a shared physical benchmark. Higher-end printers produce more consistent textures, allowing strong within-printer generalisation in classification tasks, while cross-printer generalisation remains limited by fabrication differences.

Load-bearing premise

The six patterns can be printed with enough geometric consistency across machines and filaments that differences between sensors are not swamped by fabrication noise.

Editorial extensions

If this is right

  • Any tactile sensor can now be tested against the same physical surfaces instead of sensor-specific recordings.
  • Researchers can reproduce the benchmark on their own printers and compare results directly with published data.
  • Print quality, especially peak sharpness and stringing, must be controlled because it directly sets the variance seen by the sensor.
  • Neural networks achieve high accuracy when trained and tested on prints from the same machine.
  • Cross-printer testing reveals the current limits of geometric consistency in consumer 3D printing for tactile work.

Reading between the lines

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

  • Standardised physical benchmarks could replace the current practice of publishing only sensor-specific data.
  • Future datasets might include calibration steps to normalise for printer-to-printer differences.
  • Extending the parametric family to include more complex or stochastic textures would test robustness of the benchmark further.
  • The same parametric-print approach could be applied to other sensing modalities that require reproducible physical stimuli.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

1 major / 2 minor

Summary. The paper introduces a dataset of six parametrically generated 3D-printable surface patterns (sine-wave and Fourier combinations) intended for tactile sensor evaluation. It fabricates these across three printers and multiple filaments, measures reproducibility via an optical TacTip sensor under controlled contact, reports that print quality (peak sharpness, stringing) affects tactile variance with higher-end printers yielding more consistent results, and shows strong within-printer but poor cross-printer generalization in neural network and PCA classification experiments. The work claims to provide the first openly available physically reproducible benchmark for fair tactile sensor comparisons.

Significance. If the parametric patterns can be fabricated with geometric fidelity sufficient for cross-setup comparisons, the open dataset would address a clear gap in tactile sensing by enabling standardized, reproducible testing that existing object-based datasets cannot provide. The parametric design and multi-printer evaluation are positive steps toward reproducibility. However, the internal finding of challenging cross-printer generalization due to geometric inconsistencies limits the claimed utility for fair, printer-independent comparisons.

major comments (1)
  1. [Abstract] Abstract: The central claim that the dataset 'establishes the first openly available, physically reproducible 3D-printed texture benchmark, providing a foundation for fair comparison of tactile sensors' is directly contradicted by the reported result that 'cross-printer generalisation remains challenging due to geometric inconsistencies.' This tension is load-bearing for the motivation, as the abstract itself notes that print quality affects tactile variance and higher-end printers produce more consistent signatures, indicating the benchmark may only support within-printer comparisons rather than the asserted fair cross-sensor use.
minor comments (2)
  1. The evaluation lacks reported error bars, standard deviations, or quantitative variance metrics for the tactile signatures across printers/filaments, which are needed to assess the claimed reproducibility and classification performance.
  2. Full methodological details on contact conditions, sensor calibration, and exact parametric equations for the six patterns are not provided in sufficient detail to allow independent reproduction of the dataset.

Simulated Author's Rebuttal

1 responses · 0 unresolved

We thank the referee for their constructive feedback highlighting the tension between our abstract claims and experimental results on cross-printer generalization. We address this major comment below and will revise the manuscript accordingly to ensure claims accurately reflect the findings.

read point-by-point responses

  1. Referee: [Abstract] Abstract: The central claim that the dataset 'establishes the first openly available, physically reproducible 3D-printed texture benchmark, providing a foundation for fair comparison of tactile sensors' is directly contradicted by the reported result that 'cross-printer generalisation remains challenging due to geometric inconsistencies.' This tension is load-bearing for the motivation, as the abstract itself notes that print quality affects tactile variance and higher-end printers produce more consistent signatures, indicating the benchmark may only support within-printer comparisons rather than the asserted fair cross-sensor use.

    Authors: We agree that the abstract phrasing risks overstating the scope of 'fair comparison' by not sufficiently foregrounding the printer-dependent geometric variations documented in our results. The dataset is the first openly available collection of parametrically defined, 3D-printable textures paired with reference tactile measurements, enabling reproducible testing when the same fabrication parameters and printer class are used. However, the experiments correctly show that cross-printer generalization is limited by inconsistencies in peak sharpness and stringing. We will revise the abstract to state that the benchmark provides a foundation for fair comparisons under matched fabrication conditions, while explicitly noting the observed printer-specific effects as a central finding that users must account for. This adjustment aligns the claim with the evidence without altering the core contribution of an open, parametric texture set. revision: yes

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity: empirical dataset creation and measurement study

full rationale

The paper introduces a 3D-printable texture dataset consisting of parametrically generated patterns and evaluates reproducibility via physical printing, optical sensor measurements, and classification experiments. No derivations, equations, fitted parameters presented as predictions, or self-referential claims appear in the provided text. The central claim rests on experimental results rather than any chain that reduces to its own inputs by construction. No self-citations are invoked as load-bearing uniqueness theorems or ansatzes. This is a standard empirical contribution with no detectable circularity in its logic.

Assumptions & free parameters 0 free parameters · 1 assumptions · 0 invented entities

The paper relies on standard mathematical functions (sine waves, Fourier series) and existing 3D printing technology. No new free parameters, axioms beyond domain assumptions, or invented entities are introduced.

assumptions (1)
  • domain assumption Combinations of sine-wave and Fourier-based functions produce surface patterns with controllable spatial frequency, amplitude, and directional structure suitable for tactile sensing.
    Invoked in the description of the six texture patterns.

how reviews work

0 comments
Cite this review

Pith. "Pith review of A 3D-Printable Dataset for Fair Testing and Comparisons of Tactile Sensors." pith.science (2026). https://pith.science/paper/T33L7YKW

@misc{pith2026260625886,
  author       = {Pith},
  title        = {Pith review of: A 3D-Printable Dataset for Fair Testing and Comparisons of Tactile Sensors},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/T33L7YKW}},
  note         = {Machine review of arXiv:2606.25886}
}
read the original abstract

Existing texture datasets for tactile sensing primarily consist of sensor readings from a specific sensor interacting with available surfaces/objects rather than describing the textures themselves, limiting fair comparison between tactile sensors and hindering reproducible research. In this work, we introduce a 3D-printable dataset of mathematically defined textures designed to be fabricated reliably across different printers and filament types. The dataset consists of six parametrically generated surface patterns derived from combinations of sine-wave and Fourier-based functions, giving controlled variation in spatial frequency, amplitude, and directional structure. We evaluate the reproducibility of these textures across three popular 3D printers and multiple filament types by measuring variance in images captured using an optical TacTip sensor under controlled contact conditions. Our results show that print quality, particularly peak sharpness and stringing, affects tactile variance, with higher-end printers producing significantly more consistent signatures. Classification experiments using neural networks and PCA-based models further demonstrate that high-quality prints support strong within-printer generalisation, while cross-printer generalisation remains challenging due to geometric inconsistencies. This work establishes the first openly available, physically reproducible 3D-printed texture benchmark, providing a foundation for fair comparison of tactile sensors.

Figures

Figures reproduced from arXiv: 2606.25886 by the authors.

Figure 1
Figure 1. The 3D printable textures generated from the equations listed. We can see that there is a [PITH_FULL_IMAGE:figures/full_fig_p004_1.png] view at source ↗
Figure 2
Figure 2. Example wireframe of the STL file for texture 1 generated by equation 1. We used step size [PITH_FULL_IMAGE:figures/full_fig_p005_2.png] view at source ↗
Figure 3
Figure 3. A diagram to show an example block, and the positions that the readings have been taken [PITH_FULL_IMAGE:figures/full_fig_p006_3.png] view at source ↗
Figures from the paper (5 more)
Figure 4
Figure 4. Figure 4: Rig with new texture housing unit attached. Each texture slots into the unit, allowing for [PITH_FULL_IMAGE:figures/full_fig_p006_4.png]
Figure 5
Figure 5. Figure 5: The 3D printable textures printed in PLA- on a Bambu printer. [PITH_FULL_IMAGE:figures/full_fig_p008_5.png]
Figure 6
Figure 6. Figure 6: Examples of the prints. The black filament (close up shown in [PITH_FULL_IMAGE:figures/full_fig_p009_6.png]
Figure 7
Figure 7. Figure 7: The variance between pixels for each texture, based on the Ender PLA- dataset, with [PITH_FULL_IMAGE:figures/full_fig_p011_7.png]
Figure 8
Figure 8. Figure 8: Confusion matrix for the random forest classifier trained on the Bambu dataset and evaluated [PITH_FULL_IMAGE:figures/full_fig_p013_8.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

20 extracted references · 5 canonical work pages

  1. [1]

    N. F. Lepora,Soft biomimetic optical tactile sensing with the TacTip: A review, IEEE Sensors Journal, vol. 21, no. 19, pp. 21131–21143, 2021

    2021

  2. [3]

    Ward-Cherrier, Benjamin and Pestell, Nicholas and Cramphorn, Luke and Winstone, Ben- 13 jamin and Giannaccini, Maria Elena and Rossiter, Jonathan and Lepora, Nathan F. 2018. The tactip family: Soft optical tactile sensors with 3d-printed biomimetic morphologies. Soft robotics

    2018

  3. [4]

    Winstone, Benjamin and Griffiths, Gareth and Melhuish, Chris and Pipe, Tony and Rossiter, Jonathan. 2012. TACTIP—Tactile fingertip device, challenges in reduction of size to ready for robot hand integration. 2012 IEEE International Conference on Robotics and Biomimetics ROBIO

    2012

  4. [5]

    James, Jasper Wollaston and Pestell, Nicholas and Lepora, Nathan F. 2018. Slip detection with a biomimetic tactile sensor. IEEE Robotics and Automation Letters

    2018

  5. [6]

    Meribout, N

    M. Meribout, N. A. Takele, O. Derege, N. Rifiki, M. El Khalil, V. Tiwari, and J. Zhong, ”Tactile sensors: A review,”Measurement, p. 115332, 2024

    2024

  6. [7]

    D. R. Shepherd, P. Husbands, A. Philippides, and C. Johnson,Texture and Friction Classifica- tion: Optical TacTip vs. Vibrational Piezoeletric and Accelerometer Tactile Sensors, Sensors, vol. 25, no. 16, art. 4971, 2025, doi: 10.3390/s25164971

    work page doi:10.3390/s25164971 2025

  7. [8]

    B. M. R. Lima, V. N. S. S. Danyamraju, T. E. A. de Oliveira, and V. P. da Fonseca,A multimodal tactile dataset for dynamic texture classification, Data in Brief, vol. 50, p. 109590, 2023

    2023

  8. [9]

    Marzani, S

    M. Marzani, S. Khatibi, R. Masinjila, V. P. da Fonseca, and T. E. A. de Oliveira,A dataset for tactile textures on uneven surfaces collected using a BioIn-Tacto sensing module, Data in Brief, p. 111312, 2025

    2025

Show all 20 references
  1. [10]

    Zhong, A

    S. Zhong, A. Albini, P. Maiolino, and I. Posner,TactGen: Tactile Sensory Data Generation via Zero-Shot Sim-to-Real Transfer, IEEE Transactions on Robotics, 2024

    2024

  2. [11]

    Gao, Y.-Y

    R. Gao, Y.-Y. Chang, S. Mall, L. Fei-Fei, and J. Wu,Objectfolder: A dataset of objects with implicit visual, auditory, and tactile representations, arXiv preprint arXiv:2109.07991, 2021

    2021

  3. [12]

    Cepri´ a-Bernal and A

    J. Cepri´ a-Bernal and A. P´ erez-Gonz´ alez,Dataset of tactile signatures of the human right hand in twenty-one activities of daily living using a high spatial resolution pressure sensor, Sensors, vol. 21, no. 8, p. 2594, 2021

    2021

  4. [13]

    Z. Gao, L. Chang, B. Ren, J. Han, and J. Li,Enhanced braille recognition based on piezoresis- tive and piezoelectric dual-mode tactile sensors, Sensors and Actuators A: Physical, vol. 366, p. 115000, 2024

    2024

  5. [14]

    X. Xu, N. F. Lepora, and B. Ward-Cherrier,A Neuromorphic Tactile System for Reliable Braille Reading in Noisy Environments, IEEE Robotics and Automation Letters, 2025

    2025

  6. [15]

    Lendvai, I

    L. Lendvai, I. Fekete, S. K. Jakab, G. Szarka, K. Vereb´ elyi, and B. Iv´ an,Influence of environ- mental humidity during filament storage on the structural and mechanical properties of mate- rial extrusion 3D-printed poly (lactic acid) parts, Results in Engineering, vol. 24, ...

    2024

  7. [16]

    Ward-Cherrier, N

    B. Ward-Cherrier, N. Pestell, and N. F. Lepora,NeuroTac: A neuromorphic optical tactile sensor applied to texture recognition, in Proc. IEEE Int. Conf. Robotics and Automation (ICRA), 2020, pp. 2654–2660

    2020

  8. [17]

    Tymms, D

    C. Tymms, D. Zorin, and E. P. Gardner,Tactile perception of the roughness of 3D-printed textures, Journal of Neurophysiology, vol. 119, no. 3, pp. 862–876, 2018

    2018

  9. [18]

    Cavalin and L

    P. Cavalin and L. S. Oliveira,A Review of Texture Classification Methods and Databases, in Proc. 30th SIBGRAPI Conf. Graphics, Patterns and Images Tutorials (SIBGRAPI-T), 2017, pp. 1–8, doi: 10.1109/SIBGRAPI-T.2017.10

    2017 doi

  10. [19]

    B. M. R. Lima, V. P. da Fonseca, T. E. A. de Oliveira, Q. Zhu, and E. M. Petriu,Dynamic tactile exploration for texture classification using a miniaturized multi-modal tactile sensor and machine learning, in Proc. IEEE Int. Systems Conf. (SysCon), 2020, pp. 1–7. 14

    2020

  11. [20]

    Pratap, J

    S. Pratap, J. Narayan, Y. Hatta, K. Ito, and S. M. Hazarika,From tactile signals to grasp classification: Exploring patterns with machine learning, in Proc. IEEE Int. Conf. Interdisci- plinary Approaches in Technology and Management for Social Innovation (IATMSI), vol. 2, 2024...

    2024

  12. [21]

    D. R. Shepherd,3D Printable Tactile Dataset, University of Sussex, 2025. Available: https://doi.org/10.25377/sussex.30256453 15

    2025 doi

Pith tools

Reviewed June 25, 2026 · model on record in the stance chip above.