pith. sign in

arxiv: 2604.07423 · v1 · submitted 2026-04-08 · 💻 cs.RO · cs.LG

OpenPRC: A Unified Open-Source Framework for Physics-to-Task Evaluation in Physical Reservoir Computing

Yogesh Phalak , Wen Sin Lor , Apoorva Khairnar , Benjamin Jantzen , Noel Naughton , Suyi Li This is my paper

Pith reviewed 2026-05-10 17:55 UTC · model grok-4.3

classification 💻 cs.RO cs.LG
keywords physical reservoir computingopen-source frameworkHDF5 schemaphysics simulationexperimental data ingestionbenchmarkingorigami tessellationsphysics-aware optimization
0
0 comments X

The pith

OpenPRC provides a unified open-source Python framework for consistent physics-to-task evaluation in physical reservoir computing.

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

The paper introduces OpenPRC to address the fragmented workflows in physical reservoir computing by creating a single pipeline that works with both simulated and experimental data. It consists of five modules including a physics engine, data ingestion from videos, learning tools, analysis, and optimization, all tied together by a common HDF5 data schema. This setup allows researchers to simulate physical systems like origami structures or process real measurements and then evaluate them using the same methods. A reader would care if they want to develop or compare different physical computing systems without switching between incompatible tools. The framework aims to become a standard for the community.

Core claim

We present OpenPRC, an open-source Python framework that fills this gap through a schema-driven physics-to-task pipeline built around five modules: a GPU-accelerated hybrid RK4-PBD physics engine, a video-based experimental ingestion layer, a modular learning layer, information-theoretic analysis and benchmarking tools, and physics-aware optimization. A universal HDF5 schema enforces reproducibility and interoperability, allowing GPU-simulated and experimentally acquired trajectories to enter the same downstream workflow without modification.

What carries the argument

The universal HDF5 schema combined with the five modules that create an interoperable pipeline from physical dynamics to task performance.

Load-bearing premise

That the universal HDF5 schema and the five modules can faithfully represent both high-fidelity simulated trajectories and real experimental measurements from diverse physical substrates.

What would settle it

A direct comparison where the same PRC task is run on video-captured experimental data and matching GPU-simulated trajectories from the same physical system, then checking whether the downstream task accuracy, correlation diagnostics, and capacity metrics agree within expected noise levels.

Figures

Figures reproduced from arXiv: 2604.07423 by Apoorva Khairnar, Benjamin Jantzen, Noel Naughton, Suyi Li, Wen Sin Lor, Yogesh Phalak.

Figure 1
Figure 1. Figure 1: The OpenPRC pipeline. Both demlat and openprc.vision produce simulation.h5, making simulated and experimental trajectories interchangeable to all downstream modules. The optimize module closes the loop back to demlat for iterative design optimization. 5 [PITH_FULL_IMAGE:figures/full_fig_p005_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Bar-hinge element showing the four-node hinge configuration. Node [PITH_FULL_IMAGE:figures/full_fig_p007_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Versatile applications of the bar-hinge model in [PITH_FULL_IMAGE:figures/full_fig_p008_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: DEMLAT Player visualizing a Miura-ori simulation. Faces are colored by axial bar strain (cool– [PITH_FULL_IMAGE:figures/full_fig_p010_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: Output of the openprc.vision pipeline applied to a physical Miura-ori reservoir [11]. Left: SIFT keypoints detected on the reference frame, with feature indices and response circles overlaid. Right: KLT tracking in progress, showing per-feature trajectory trails and forward-backward status indicators. The status bar at the bottom encodes per-feature tracking health across the frame window. The output file … view at source ↗
Figure 6
Figure 6. Figure 6: Representative NARMA2 benchmark result for the Miura-ori reservoir. The target is generated [PITH_FULL_IMAGE:figures/full_fig_p016_6.png] view at source ↗
Figure 7
Figure 7. Figure 7: Memory-benchmark result for the Miura-ori reservoir under the truncated Dambre-style IPC [PITH_FULL_IMAGE:figures/full_fig_p018_7.png] view at source ↗
read the original abstract

Physical Reservoir Computing (PRC) leverages the intrinsic nonlinear dynamics of physical substrates, mechanical, optical, spintronic, and beyond, as fixed computational reservoirs, offering a compelling paradigm for energy-efficient and embodied machine learning. However, the practical workflow for developing and evaluating PRC systems remains fragmented: existing tools typically address only isolated parts of the pipeline, such as substrate-specific simulation, digital reservoir benchmarking, or readout training. What is missing is a unified framework that can represent both high-fidelity simulated trajectories and real experimental measurements through the same data interface, enabling reproducible evaluation, analysis, and physics-aware optimization across substrates and data sources. We present OpenPRC, an open-source Python framework that fills this gap through a schema-driven physics-to-task pipeline built around five modules: a GPU-accelerated hybrid RK4-PBD physics engine (demlat), a video-based experimental ingestion layer (openprc.vision), a modular learning layer (reservoir), information-theoretic analysis and benchmarking tools (analysis), and physics-aware optimization (optimize). A universal HDF5 schema enforces reproducibility and interoperability, allowing GPU-simulated and experimentally acquired trajectories to enter the same downstream workflow without modification. Demonstrated capabilities include simulations of Origami tessellations, video-based trajectory extraction from a physical reservoir, and a common interface for standardized PRC benchmarking, correlation diagnostics, and capacity analysis. The longer-term vision is to serve as a standardizing layer for the PRC community, compatible with external physics engines including PyBullet, PyElastica, and MERLIN.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

2 major / 2 minor

Summary. The manuscript introduces OpenPRC, an open-source Python framework for Physical Reservoir Computing (PRC) consisting of five modules (GPU-accelerated hybrid RK4-PBD physics engine demlat, video-based experimental ingestion openprc.vision, modular reservoir learning, information-theoretic analysis and benchmarking, and physics-aware optimization) connected by a universal HDF5 schema. The central claim is that this schema-driven pipeline enables simulated trajectories and real experimental measurements to enter identical downstream workflows without modification, supporting reproducible PRC evaluation, benchmarking, and optimization across substrates, with demonstrations on Origami tessellation simulations and physical reservoir video extraction; the longer-term goal is community standardization and compatibility with external engines such as PyBullet.

Significance. If the interoperability and schema claims hold, the framework would address a genuine fragmentation in PRC tooling by providing a common physics-to-task interface, which could improve reproducibility, enable direct simulation-experiment comparisons, and accelerate embodied ML research; the open-source nature and modular design are positive factors for adoption.

major comments (2)
  1. [Abstract] Abstract: the claim that 'GPU-simulated and experimentally acquired trajectories [enter] the same downstream workflow without modification' is load-bearing for the paper's contribution but is not supported by evidence. The manuscript shows separate demonstrations (Origami tessellation simulation; video-based trajectory extraction) but provides no joint example in which the reservoir, analysis, or optimize modules are applied to both data sources with zero schema-level or module-level adaptation, leaving the interoperability assertion untested.
  2. [Framework description] Framework description (HDF5 schema section): the universal HDF5 schema is presented as sufficient to represent both high-fidelity simulated trajectories and noisy experimental measurements, yet no comparative validation, metadata completeness check, or handling of differing sampling rates/noise characteristics is reported; this directly affects whether the 'without modification' pipeline actually functions for real experimental data.
minor comments (2)
  1. [Abstract] The abstract and introduction would benefit from quantitative metrics (e.g., reservoir capacity scores, training times, or benchmarking results) rather than qualitative descriptions of capabilities.
  2. [Introduction] Compatibility statements with external engines (PyBullet, PyElastica, MERLIN) should specify version ranges or interface requirements to aid reproducibility.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the constructive feedback and for recognizing the potential of OpenPRC to address fragmentation in PRC tooling. We address each major comment below with planned revisions to strengthen the evidence supporting our interoperability claims.

read point-by-point responses

  1. Referee: [Abstract] Abstract: the claim that 'GPU-simulated and experimentally acquired trajectories [enter] the same downstream workflow without modification' is load-bearing for the paper's contribution but is not supported by evidence. The manuscript shows separate demonstrations (Origami tessellation simulation; video-based trajectory extraction) but provides no joint example in which the reservoir, analysis, or optimize modules are applied to both data sources with zero schema-level or module-level adaptation, leaving the interoperability assertion untested.

    Authors: We agree that an explicit joint demonstration would provide stronger support for the central interoperability claim. While the modular design and universal HDF5 schema are intended to enable identical downstream processing, the manuscript currently demonstrates the modules on simulated and experimental data separately. In the revision, we will add a dedicated subsection with a unified end-to-end example: both GPU-simulated Origami trajectories and video-extracted experimental trajectories will be ingested via the same schema and processed through the reservoir learning, analysis, and optimize modules with no code changes. Side-by-side benchmarking results (e.g., task performance and capacity metrics) will be reported to directly test the 'without modification' assertion. revision: yes

  2. Referee: [Framework description] Framework description (HDF5 schema section): the universal HDF5 schema is presented as sufficient to represent both high-fidelity simulated trajectories and noisy experimental measurements, yet no comparative validation, metadata completeness check, or handling of differing sampling rates/noise characteristics is reported; this directly affects whether the 'without modification' pipeline actually functions for real experimental data.

    Authors: The HDF5 schema incorporates extensible metadata attributes for sampling rates, noise statistics, timestamps, and data provenance precisely to accommodate both high-fidelity simulations and noisy experimental measurements. We acknowledge, however, that the current manuscript lacks explicit comparative validation or examples of preprocessing steps for differing sampling rates and noise. In the revision, we will expand the HDF5 schema section with a new validation subsection that includes (i) metadata completeness checks, (ii) an example of rate normalization between simulated and experimental trajectories, and (iii) basic noise-handling metadata usage, thereby demonstrating that the pipeline functions without downstream modification for real data. revision: yes

Circularity Check

0 steps flagged

No circularity: software framework description with no derivations or fitted predictions

full rationale

The paper describes an open-source Python framework (OpenPRC) consisting of five modules and a universal HDF5 schema for handling simulated and experimental trajectories in Physical Reservoir Computing. No mathematical derivations, equations, parameter fittings, or predictions are present in the provided text or abstract. Claims about interoperability and unified workflows are design assertions supported by module descriptions and separate demonstrations, not reductions to self-referential inputs or self-citations. The paper is self-contained as a tool presentation and does not invoke uniqueness theorems, ansatzes, or renamed empirical patterns in a load-bearing way.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

This is a software framework paper rather than a theoretical or empirical scientific claim, so it introduces no free parameters, axioms, or invented physical entities.

pith-pipeline@v0.9.0 · 5595 in / 1273 out tokens · 49906 ms · 2026-05-10T17:55:09.525344+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Lean theorems connected to this paper

Citations machine-checked in the Pith Canon. Every link opens the source theorem in the public Lean library.

What do these tags mean?
matches
The paper's claim is directly supported by a theorem in the formal canon.
supports
The theorem supports part of the paper's argument, but the paper may add assumptions or extra steps.
extends
The paper goes beyond the formal theorem; the theorem is a base layer rather than the whole result.
uses
The paper appears to rely on the theorem as machinery.
contradicts
The paper's claim conflicts with a theorem or certificate in the canon.
unclear
Pith found a possible connection, but the passage is too broad, indirect, or ambiguous to say the theorem truly supports the claim.

Reference graph

Works this paper leans on

40 extracted references · 40 canonical work pages

  1. [1]

    echo state

    Herbert Jaeger. The “echo state” approach to analysing and training recurrent neural networks. GMD Report 148, German National Research Center for Information Technology (GMD), 2001. URL https://doi.org/10.24406/publica-fhg-291111

    work page doi:10.24406/publica-fhg-291111 2001

  2. [2]

    Real-time computing without stable states: A new framework for neural computation based on perturbations.Neural Computation, 14(11):2531–2560, 2002

    Wolfgang Maass, Thomas Natschl¨ ager, and Henry Markram. Real-time computing without stable states: A new framework for neural computation based on perturbations.Neural Computation, 14(11): 2531–2560, 2002. URLhttps://doi.org/10.1162/089976602760407955

    work page doi:10.1162/089976602760407955 2002

  3. [3]

    (28) Zhong, Y.; Tang, J.; Li, X.; Gao, B.; Qian, H.; Wu, H., Dynamic memristor - based reservoir computing for high - efficiency temporal signal processing

    Gouhei Tanaka, Toshiyuki Yamane, Jean Benoit H´ eroux, Ryosho Nakane, Naoki Kanazawa, Seiji Takeda, Hidetoshi Numata, Daiju Nakano, and Akira Hirose. Recent advances in physical reservoir computing: A review.Neural Networks, 115:100–123, 2019. URL https://doi.org/10.1016/j.neunet.2019.03.005

    work page doi:10.1016/j.neunet.2019.03.005 2019

  4. [4]

    Physical reservoir computing—an introductory perspective.Japanese Journal of Applied Physics, 59(6):060501, 2020

    Kohei Nakajima. Physical reservoir computing—an introductory perspective.Japanese Journal of Applied Physics, 59(6):060501, 2020. ISSN 1347-4065. URLhttps://doi.org/10.35848/1347-4065/ab8d4f

    work page doi:10.35848/1347-4065/ab8d4f 2020

  5. [5]

    Allwood, Matthew O

    Dan A. Allwood, Matthew O. A. Ellis, David Griffin, Thomas J. Hayward, Luca Manneschi, Mohammad F. Kh. Musameh, Simon O’Keefe, Susan Stepney, Charles Swindells, Martin A. Trefzer, Eleni Vasilaki, Guru Venkat, Ian Vidamour, and Chester Wringe. A perspective on physical reservoir computing with nanomagnetic devices.Applied Physics Letters, 122(4):040501, 20...

    work page doi:10.1063/5.0119040 2023

  6. [6]

    Physical reservoir computing with emerging electronics.Nature Electronics, 7(3):193–206, 2024

    Xiangpeng Liang, Jianshi Tang, Yanan Zhong, Bin Gao, He Qian, and Huaqiang Wu. Physical reservoir computing with emerging electronics.Nature Electronics, 7(3):193–206, 2024. ISSN 2520-1131. URL https://doi.org/10.1038/s41928-024-01133-z

    work page doi:10.1038/s41928-024-01133-z 2024

  7. [7]

    Harnessing synthetic active particles for physical reservoir computing

    Xiangzun Wang and Frank Cichos. Harnessing synthetic active particles for physical reservoir computing. Nature Communications, 15:774, 2024. URLhttps://doi.org/10.1038/s41467-024-44856-5

    work page doi:10.1038/s41467-024-44856-5 2024

  8. [8]

    Sabelhaus, Jonathan Bruce, Benjamin Schrauwen, and Vytas SunSpiral

    Ken Caluwaerts, J´ er´ emie Despraz, Atıl I¸ s¸ cen, Andrew P. Sabelhaus, Jonathan Bruce, Benjamin Schrauwen, and Vytas SunSpiral. Design and control of compliant tensegrity robots through sim- ulation and hardware validation.Journal of the Royal Society Interface, 11(98):20140520, 2014. URL https://doi.org/10.1098/rsif.2014.0520

    work page doi:10.1098/rsif.2014.0520 2014

  9. [9]

    Ex- ploiting short-term memory in soft body dynamics as a computational resource,

    K. Nakajima, T. Li, H. Hauser, and R. Pfeifer. Exploiting short-term memory in soft body dynamics as a computational resource.Journal of the Royal Society Interface, 11(100):20140437, 2014. URL https://doi.org/10.1098/rsif.2014.0437

    work page doi:10.1098/rsif.2014.0437 2014

  10. [10]

    Enhanc- ing time series data classification for structural damage detection through out-of-distribution representation learning

    Shan He and Patrick Musgrave. Physical reservoir computing on a soft bio-inspired swimmer.Neural Networks, 181:106766, 2025. URLhttps://doi.org/10.1016/j.neunet.2024.106766

    work page doi:10.1016/j.neunet.2024.106766 2025

  11. [11]

    Physical reservoir computing with origami and its application to robotic crawling.Scientific Reports, 11:13002, 2021

    Priyanka Bhovad and Suyi Li. Physical reservoir computing with origami and its application to robotic crawling.Scientific Reports, 11:13002, 2021. URLhttps://doi.org/10.1038/s41598-021-92257-1. 21

    work page doi:10.1038/s41598-021-92257-1 2021

  12. [12]

    Cellular automata inspired multistable origami metamaterials for mechanical learning.Advanced Science, 10(34):e2305146, 2023

    Zuolin Liu, Hongbin Fang, Jian Xu, and Kon-Well Wang. Cellular automata inspired multistable origami metamaterials for mechanical learning.Advanced Science, 10(34):e2305146, 2023. URL https: //doi.org/10.1002/advs.202305146

    work page doi:10.1002/advs.202305146 2023

  13. [13]

    Jean-Charles Coulombe, Mark C. A. York, and Julien Sylvestre. Computing with networks of nonlinear mechanical oscillators.PLOS ONE, 12(6):e0178663, 2017. URL https://doi.org/10.1371/journal. pone.0178663

    work page doi:10.1371/journal 2017

  14. [14]

    Dambre, D

    Joni Dambre, David Verstraeten, Benjamin Schrauwen, and Serge Massar. Information processing capacity of dynamical systems.Scientific Reports, 2:514, 2012. URLhttps://doi.org/10.1038/srep00514

    work page doi:10.1038/srep00514 2012

  15. [15]

    Ke Liu and Glaucio H. Paulino. Merlin: A matlab implementation to capture highly nonlinear behavior of non-rigid origami. InProceedings of IASS Annual Symposia, pages 1–10. International Association for Shell and Spatial Structures (IASS), 2016. URL https://paulino.princeton.edu/conferences/ papers/16Liu_merlin.pdf

    work page 2016

  16. [16]

    Ke Liu and Glaucio H. Paulino. Nonlinear mechanics of non-rigid origami: An efficient computational approach.Proceedings of the Royal Society A: Mathematical, Physical and Engineering Sciences, 473 (2206):20170348, 2017. URLhttps://doi.org/10.1098/rspa.2017.0348

    work page doi:10.1098/rspa.2017.0348 2017

  17. [17]

    Ke Liu and Glaucio H. Paulino. Highly efficient nonlinear structural analysis of origami assemblages using the merlin2 software. In Robert J. Lang, Mark Bolitho, and Zhong You, editors,Origami7: Proceedings of the 7th International Meeting on Origami in Science, Mathematics and Education, pages 1167–1182. Tarquin Publications, St Albans, UK, 2018. URL htt...

    work page 2018

  18. [18]

    Overcoming catastrophic forgetting in neural networks

    Evgueni T. Filipov, Tomohiro Tachi, and Glaucio H. Paulino. Origami tubes assembled into stiff, yet reconfigurable structures and metamaterials.Proceedings of the National Academy of Sciences of the United States of America, 112(40):12321–12326, 2015. URL https://doi.org/10.1073/pnas. 1509465112

    work page doi:10.1073/pnas 2015

  19. [19]

    Yi Zhu, Mark Schenk, and Evgueni T. Filipov. A review on origami simulations: From kinematics, to mechanics, toward multiphysics.Applied Mechanics Reviews, 74(3):030801, 2022. ISSN 0003-6900. URL https://doi.org/10.1115/1.4055031

    work page doi:10.1115/1.4055031 2022

  20. [20]

    Falanga, S

    Noel Naughton, Jiarui Sun, Arman Tekinalp, Tejaswin Parthasarathy, Girish Chowdhary, and Mattia Gazzola. Elastica: A compliant mechanics environment for soft robotic control.IEEE Robotics and Automation Letters, 6(2):3389–3396, 2021. ISSN 2377-3766. URL https://doi.org/10.1109/LRA. 2021.3063698

    work page doi:10.1109/lra 2021

  21. [21]

    Gazzola, L

    M. Gazzola, L. H. Dudte, A. G. McCormick, and L. Mahadevan. Forward and inverse problems in the mechanics of soft filaments.Royal Society Open Science, 5(6):171628, 2018. URL https: //doi.org/10.1098/rsos.171628

    work page doi:10.1098/rsos.171628 2018

  22. [22]

    ReservoirPy: An efficient and user-friendly library to design echo state networks

    Nathan Trouvain, Luca Pedrelli, Thanh Trung Dinh, and Xavier Hinaut. ReservoirPy: An efficient and user-friendly library to design echo state networks. InArtificial Neural Networks and Machine Learning – ICANN 2020, Lecture Notes in Computer Science, pages 494–505. Springer International Publishing,

    work page 2020

  23. [23]

    URLhttps://doi.org/10.1007/978-3-030-61616-8_40

    work page doi:10.1007/978-3-030-61616-8_40

  24. [24]

    Williams, Dima Tretiak, Steven L

    Jan P. Williams, Dima Tretiak, Steven L. Brunton, J. Nathan Kutz, and Krithika Manohar. OpenReser- voirComputing: GPU-accelerated reservoir computing in jax, 2026. URL https://arxiv.org/abs/ 2603.14802

    work page arXiv 2026

  25. [25]

    Stenning, Jack C

    Harry Youel, Daniel Prestwood, Oscar Lee, Tianyi Wei, Kilian D. Stenning, Jack C. Gartside, Will R. Branford, Karin Everschor-Sitte, and Hidekazu Kurebayashi. PRCpy: A Python package for processing of physical reservoir computing, 2024. URLhttps://doi.org/10.48550/ARXIV.2410.18356

    work page doi:10.48550/arxiv.2410.18356 2024

  26. [26]

    Reservoir computing approaches to recurrent neural network 22 training.Computer Science Review, 3(3):127–149, 2009

    Mantas Lukoˇ seviˇ cius and Herbert Jaeger. Reservoir computing approaches to recurrent neural network 22 training.Computer Science Review, 3(3):127–149, 2009. ISSN 1574-0137. URL https://doi.org/10. 1016/j.cosrev.2009.03.005

    work page 2009

  27. [27]

    Short term memory in echo state networks

    Herbert Jaeger. Short term memory in echo state networks. GMD Report 152, German Na- tional Research Center for Information Technology (GMD), 2001. URL https://doi.org/10.24406/ publica-fhg-291107

    work page 2001

  28. [28]

    Position based dynamics , year =

    Matthias M¨ uller, Bruno Heidelberger, Marcus Hennix, and John Ratcliff. Position based dynamics. Journal of Visual Communication and Image Representation, 18(2):109–118, 2007. URL https://doi. org/10.1016/j.jvcir.2007.01.005

    work page doi:10.1016/j.jvcir.2007.01.005 2007

  29. [29]

    Piviz: High-performance scientific visualization, 2026

    Yogesh Phalak. Piviz: High-performance scientific visualization, 2026. URL https://github.com/ PhalcoAi/PiViz

    work page 2026

  30. [30]

    David G. Lowe. Distinctive image features from scale-invariant keypoints.International Journal of Computer Vision, 60(2):91–110, 2004. URL https://doi.org/10.1023/B:VISI.0000029664.99615. 94

    work page doi:10.1023/b:visi.0000029664.99615 2004

  31. [31]

    ORB: An efficient alternative to SIFT or SURF

    Ethan Rublee, Vincent Rabaud, Kurt Konolige, and Gary Bradski. ORB: An efficient alternative to SIFT or SURF. InProceedings of the IEEE International Conference on Computer Vision (ICCV), pages 2564–2571, 2011. URLhttps://doi.org/10.1109/ICCV.2011.6126544

    work page doi:10.1109/iccv.2011.6126544 2011

  32. [32]

    Fast explicit diffusion for accelerated features in nonlinear scale spaces

    Pablo Fern´ andez Alcantarilla, Jes´ us Nuevo, and Adrien Bartoli. Fast explicit diffusion for accelerated features in nonlinear scale spaces. InProceedings of the British Machine Vision Conference (BMVC), pages 13.1–13.11, 2013. URLhttps://doi.org/10.5244/C.27.13

    work page doi:10.5244/c.27.13 2013

  33. [33]

    Lucas and Takeo Kanade

    Bruce D. Lucas and Takeo Kanade. An iterative image registration technique with an application to stereo vision. InProceedings of the 7th International Joint Conference on Artificial Intelligence (IJCAI), pages 674–679, 1981

    work page 1981

  34. [34]

    Pyramidal implementation of the Lucas–Kanade feature tracker

    Jean-Yves Bouguet. Pyramidal implementation of the Lucas–Kanade feature tracker. Technical report, Intel Corporation, Microprocessor Research Labs, 2001

    work page 2001

  35. [35]

    Forward-backward error: Automatic detection of tracking failures

    Zdenek Kalal, Krystian Mikolajczyk, and Jiri Matas. Forward-backward error: Automatic detection of tracking failures. InProceedings of the 20th International Conference on Pattern Recognition (ICPR), pages 2756–2759, 2010. URLhttps://doi.org/10.1109/ICPR.2010.675

    work page doi:10.1109/icpr.2010.675 2010

  36. [36]

    A Flexible New Technique for Camera Calibration.IEEE Transactions on Pattern Analysis and Machine Intelligence2000,22, 1330–1334

    Zhengyou Zhang. A flexible new technique for camera calibration.IEEE Transactions on Pattern Analysis and Machine Intelligence, 22(11):1330–1334, 2000. URLhttps://doi.org/10.1109/34.888718

    work page doi:10.1109/34.888718 2000

  37. [37]

    The OpenCV library.Dr

    Gary Bradski. The OpenCV library.Dr. Dobb’s Journal of Software Tools, 25(11):120–125, 2000

    work page 2000

  38. [38]

    Sz´ ekely, Maria L

    G´ abor J. Sz´ ekely, Maria L. Rizzo, and Nail K. Bakirov. Measuring and testing dependence by cor- relation of distances.Annals of Statistics, 35(6):2769–2794, 2007. URL https://doi.org/10.1214/ 009053607000000505

    work page 2007

  39. [39]

    Measuring statistical dependence with Hilbert-Schmidt norms

    Arthur Gretton, Olivier Bousquet, Alex Smola, and Bernhard Sch¨ olkopf. Measuring statistical dependence with Hilbert-Schmidt norms. InProceedings of the 16th International Conference on Algorithmic Learning Theory (ALT), pages 63–77, 2005. URLhttps://doi.org/10.1007/11564089_7

    work page doi:10.1007/11564089_7 2005

  40. [40]

    The effective rank: A measure of effective dimensionality

    Olivier Roy and Martin Vetterli. The effective rank: A measure of effective dimensionality. InProceedings of the 15th European Signal Processing Conference (EUSIPCO), pages 606–610, 2007. 23

    work page 2007