Skip to main content

Thank you for visiting nature.com. You are using a browser version with limited support for CSS. To obtain the best experience, we recommend you use a more up to date browser (or turn off compatibility mode in Internet Explorer). In the meantime, to ensure continued support, we are displaying the site without styles and JavaScript.

  • Correspondence
  • Published:

JDLL: a library to run deep learning models on Java bioimage informatics platforms

Nature Methods volume 21, pages 7–8 (2024)Cite this article

Subjects

The advancements in artificial intelligence (AI) technology over the past decade have been a breakthrough in imaging for life sciences, paving the way for novel methods in image restoration1, reconstruction2 and segmentation3. However, the wide adoption of deep learning (DL) techniques by end users in bioimage analysis is hindered by the complexity of their deployment. These techniques stem from a variety of rapidly evolving frameworks (for example, TensorFlow 1 or 2, PyTorch) that come with distinct and often conflicting setups, which can discourage even proficient developers. This has led to integration difficulties or even absence in mainstream bioimage informatics platforms such as ImageJ, Icy and Fiji, many of which are primarily developed in Java.

The flexibility and adaptability of these components are pivotal in facilitating the effortless installation and utilization of the full spectrum of DL models prevalent in the life sciences. JDLL is streamlined, carrying minimal dependencies — namely, ImgLib2 (ref. 4) and specific libraries for processing JSON and YAML files, commonly used for DL model specifications. It is built to be compatible with existing and potential future DL frameworks, leveraging ImgLib2 to create a framework-agnostic tensor implementation. Consequently, applications built on JDLL will not require engine-specific code. JDLL offers a universal, singular implementation compatible with all engines across their various versions. Its cutting-edge management of DL engines empowers it to operate multiple models within a single program, even if they were trained on normally incompatible DL engines. For example, JDLL can effortlessly sequence a style-transfer model operating on Tensorflow 2 with an instance segmentation model running on Tensorflow 1 within a single script.

This is a preview of subscription content, access via your institution

Access options

Buy this article

  • Purchase on Springer Link
  • Instant access to full article PDF
Buy now

Prices may be subject to local taxes which are calculated during checkout

Fig. 1: The JDLL architecture.

Code availability

The source code, documentation, tutorials and examples implementing JDLL can be found at https://github.com/bioimage-io/JDLL. JDLL is made available under the open-source Apache software license.

References

  1. Weigert, M. et al. Nat. Methods 15, 1090–1097 (2018).

    Article  CAS  PubMed  Google Scholar 

  2. Belthangady, C. & Royer, L. A. Nat. Methods 16, 1215–1225 (2019).

    Article  CAS  PubMed  Google Scholar 

  3. Moen, E. et al. Nat. Methods 16, 1233–1246 (2019).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  4. Pietzsch, T., Preibisch, S., Tomancak, P. & Saalfeld, S. Bioinformatics 28, 3009–3011 (2012).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  5. Ouyang, W. et al. Preprint at bioRxiv https://doi.org/10.1101/2022.06.07.495102 (2022).

  6. de Chaumont, F. et al. Nat. Methods 9, 690–696 (2012).

    Article  PubMed  Google Scholar 

  7. Gómez-de-Mariscal, E. et al. Nat. Methods 18, 1192–1195 (2021).

    Article  PubMed  Google Scholar 

  8. von Chamier, L. et al. Nat. Commun. 12, 2276 (2021).

    Article  Google Scholar 

  9. Ouyang, W., Mueller, F., Hjelmare, M., Lundberg, E. & Zimmer, C. Nat. Methods 16, 1199–1200 (2019).

    Article  CAS  PubMed  Google Scholar 

Download references

Acknowledgements

This work has been partially supported by the Agence Nationale de la Recherche through the LabEx IBEID (ANR-10-LABX-62-IBEID), the Institut Carnot Pasteur Microbes & Santé (ANR 16 CARN 0023-01), the programs PIA INCEPTION (ANR-16-CONV-0005) and France-BioImaging (ANR-10-INBS-04); by DIM ELICIT Région Ile-de-France; by the European Commission through the H2020-FET-OPEN-2018–2019-2020-01 grant no. 862840 (“FREE@POC”) (J.-C.O.-M.); by additional internal funding from the Bioimage Analysis unit and the Institut Pasteur (J.-C.O.-M. and J.-Y.T.); by the Ministerio de Ciencia, Innovación y Universidades, Agencia Estatal de Investigación, under grant PID2019-109820RB-I00, MCIN / AEI / 10.13039/501100011033/, co-financed by European Regional Development Fund (ERDF), “A way of making Europe,” and the European Commission through the Horizon Europe program (AI4LIFE project, grant agreement 101057970-AI4LIFE) (A.M.-B.); and by Fundação Calouste Gulbenkian and EMBO Postdoctoral Fellowship (EMBO ALTF 174-2022) (E.G.M.). Funded by the European Union. Views and opinions expressed are however those of the authors only and do not necessarily reflect those of the European Union. Neither the European Union nor the granting authority can be held responsible for them.

Author information

Authors and Affiliations

  1. Bioimage Analysis Unit, Institut Pasteur, Université Paris Cité, Paris, France

    Carlos García López de Haro, Stéphane Dallongeville, Thomas Musset & Jean-Christophe Olivo-Marin

  2. CNRS UMR 3691, Institut Pasteur, Paris, France

    Stéphane Dallongeville, Thomas Musset & Jean-Christophe Olivo-Marin

  3. Instituto Gulbenkian de Ciência, Lisbon, Portugal

    Estibaliz Gómez-de-Mariscal

  4. Biomedical Imaging Group and Center for Imaging, Ecole Polytechnique de Lausanne (EPFL), Lausanne, Switzerland

    Daniel Sage

  5. Science for Life Laboratory, KTH Royal Institute of Technology, Stockholm, Sweden

    Wei Ouyang

  6. Biomedical Sciences and Engineering Laboratory, Universidad Carlos III de Madrid, Leganés, Spain

    Arrate Muñoz-Barrutia

  7. Image Analysis Hub, Institut Pasteur, Université Paris Cité, Paris, France

    Jean-Yves Tinevez

Authors
  1. Carlos García López de Haro
    View author publications

    You can also search for this author in PubMed Google Scholar

  2. Stéphane Dallongeville
    View author publications

    You can also search for this author in PubMed Google Scholar

  3. Thomas Musset
    View author publications

    You can also search for this author in PubMed Google Scholar

  4. Estibaliz Gómez-de-Mariscal
    View author publications

    You can also search for this author in PubMed Google Scholar

  5. Daniel Sage
    View author publications

    You can also search for this author in PubMed Google Scholar

  6. Wei Ouyang
    View author publications

    You can also search for this author in PubMed Google Scholar

  7. Arrate Muñoz-Barrutia
    View author publications

    You can also search for this author in PubMed Google Scholar

  8. Jean-Yves Tinevez
    View author publications

    You can also search for this author in PubMed Google Scholar

  9. Jean-Christophe Olivo-Marin
    View author publications

    You can also search for this author in PubMed Google Scholar

Contributions

Code concept and design were done by C.G.L.d.H., J.-Y.T., S.D. and J.-C.O.-M., with contributions from E.G.d.M., D.S., W.O. and A.M.-B. JDLL coding development and implementation was done by C.G.L.d.H., J.-Y.T., T.M. and S.D.; manuscript organizing and writing by C.G.L.d.H., J.-Y.T., A.M.-B. and J.-C.O.-M. with all authors contributing comments and revisions; funding and project administration by J.-Y.T. and J.-C.O.-M.

Corresponding authors

Correspondence to Jean-Yves Tinevez or Jean-Christophe Olivo-Marin.

Ethics declarations

Competing interests

The authors declare no competing interests.

Peer review

Peer review information

Nature Methods thanks David Barry and Edward Evans, III for their contribution to the peer review of this work.

Rights and permissions

Reprints and permissions

About this article

Check for updates. Verify currency and authenticity via CrossMark

Cite this article

García López de Haro, C., Dallongeville, S., Musset, T. et al. JDLL: a library to run deep learning models on Java bioimage informatics platforms. Nat Methods 21, 7–8 (2024). https://doi.org/10.1038/s41592-023-02129-x

Download citation

  • Published:

  • Issue Date:

  • DOI: https://doi.org/10.1038/s41592-023-02129-x

Search

Advanced search

Quick links

Nature Briefing AI and Robotics

Sign up for the Nature Briefing: AI and Robotics newsletter — what matters in AI and robotics research, free to your inbox weekly.

Get the most important science stories of the day, free in your inbox. Sign up for Nature Briefing: AI and Robotics