Centralised/Online image processing solutions/platforms

Here a post to reference new online platforms for multi-modal (several programing languages) and centralised image processing, usually on servers.
The motivation behind those platforms are first that there are many tools for image processing that cannot be easily combined into a single framework.
The second motivation is that the distribution of workflow, scripts… to end users can be much cumbersome when it comes to dealing with dependencies… while it is quite easy to drop some files on a web based platform (can be time consuming though).
Fiji has the advantage of the update site mechanism that makes this process easier but not every tool offer this functionality.

So I listed below a list of initiatives both open-source and commercial, that offers alternative ways to distribute image processing solutions to end users, either via a web platform and/or a remote/local centralized computational unit (ex:server).
For the solutions relying on remote servers, one bottleneck is the upload of heavy image data to those platform. If a local in-house server can be setup for such platform/solution, this would improve the transfer time, additionally ensuring that the data
are not leaving the institution which can be of interest for confidentiality issues.

I mention some people involved in those projects, feel free to contact them privately or open another forum thread if your question is related to a given solution.
If I miss some alternatives, or if you want to edit the content feel free to contribute to this post :wink:

  • “Galaxy Image” (open source)

    Galaxy is an open-source data processing framework that is popular is the field of biosequence analysis (genomics, proteomics…).
    Recently, a proof-of-concept for an image processing version of galaxy was realised by Jerome Mutterer (@jerome.mutterer) from the IBMC in Strasbourg.
    It relies on a linux server that hosts a galaxy instance, accessible via a web browser.
    Images can be uploaded on the resulting platform to run some image processing workflow.
    Workflows are made by a succession of tools, and can be designed graphically like for Knime.
    Custom “tools” can be added in a modular way by setting a configuration/command file in the galaxy filesystem.
    This configuration/command file contains the input/ouput definition (including GUI elements) as well as the instructions defining which software and script to call (Fiji, python…)
    The software (fiji, python…) associated to a tool resides on the server hosting the galaxy instance.
    This allows to have a unique installation of the external software with the proper dependencies, versions, configurations…

  • IDR (Image Data Ressource) Virtual Analysis Environment (free)

    The image data repository allows to upload biological images, share and browse various image datasets via an OMERO interface.
    It also offers the possibility to do some image processing with Jupyter notebooks.
    The basic access only allows limited storages and operations but higher storage and ressources can be requested using a GitHub identifiant.

  • ImJoy (open source) – Institut Pasteur, Paris

    This is a web plateform for image processing, that can take advantages of ressources from the local machine used to access the platfrom or from a cluster.
    In addition to the library of tools available, custom tools can be written in various languages, and a local python instance can also be connected.
    In active developpement by @oeway

  • APEER (free) – Zeiss

    This is the Zeiss web platform for image processing that allows to graphically build workflows from a list of tools relying on various languages/libraries.
    Several tools are available on the platform, and additional custom tools can be added as docker images.
    Future releases would leave the possibility to use a local server rather than the current central Zeiss server. Additionally the platform will be accessible via a local Zen blue instance, too.
    Contacts: @rkirmse, @sebi06

  • KNIME Server (commercial, free license for academic institutions) – KNIME

    The open source platfrom for data processing Knime also has a commercial extension that allows to use a server as a remote “workflow repository”.
    In this way end users can run workflows via a web page with a graphical user interface without directly interacting with the workflow.
    This option is limited to knime workflow execution, but Knime has a support for python and R extensions, while external softwares can be called within a knime workflow via the external tool node.

  • Huygens Remote Manager (HRM) (commercial) - SVI

    HRM provides a web interface for the Huygens software (while HRM itself is open-source, the Huygens software requires a license). It’s mainly used for deconvolution of image data, but also provides some analysis options, e.g. colocalization analysis.

  • Remote Desktop

    This solution consists in having a central machine that can be accessed by any workstation that belongs to the network, or just connected to the internet in some cases but the speed migth be lower.
    As the name suggests, the users get access to a “desktop session” similar to their usual operating system and can benefit from the resources and storage of the remote machine.
    This allows to have separate workspaces for the users, but also to define some centralised resources (large storage unit, some program…) that would be shared by every users.
    For instance, a central knime installation can be used with all the necessary packages and parallel instances of this KNIME installation can be run by different users in their remote desktop environnements.
    Only the knime workspaces are user-specific while the KNIME installation is shared.
    This solution is used by Acquifer for instance with their HIVE system that offers a centralised storage and workstation.
    I am currently doing my PhD on image processing there, so you can contact me (@LThomas) for more information.


Thanks @LThomas for this nice summary! I took the liberty to edit your post a bit to make it more readable, and also added a paragraph about Huygens Remote Manager.


With the latest release of Zen Blue 2.6 one can now use the APEER connector to start workflows and retrieve results directly from within the Zen software.

1 Like

I can’t edit my initial post anymore but here is an additional item for the list

  • Biaflow (by NEUBIAS)

This platform is designed for the benchmarking of different algorithm implementations for a given image processing problem (e.g. comparing ImageJ and CellProfiler for cell nuclei segmentation) using annotated reference datasets (either simulated images, or real images manually annotated) available on the platform.
The platform relies on Docker images to run the algorithms in the appropriate environment.
Accessible at https://biaflows.neubias.org with Guest access (no right for job execution).
user : guest
pass : guest
Contributions with algorithm or datasets are welcomed !

Thanks @LThomas for mentioning BIAFLOWS. It is now being much better documented (http://biaflows-doc.neubias.org/) and with more content and a redesigned user-friendly interface. See also our video on the front page of https://biaflows.neubias.org/
Please note it is fully open-source.