Skip to content

[FEATURE] Remote server support for Cellpose GUI inference, training, and annotation #1424

Description

@bobqianic

Before you fill out this form:

Did you review the FAQ?

Did you look through previous (open AND closed) issues posted on GH?

Now fill this form out completely:

Is your feature request related to a problem? Please describe.

Yes. Many users work on laptops or lower-power local machines, which are convenient for annotation and review but not ideal for heavy model inference or training. Running Cellpose models locally can be slow or impractical when GPU resources are limited.

It would be very helpful if the Cellpose GUI could connect to a remote server so that inference and training run on the server, while the user continues annotating and interacting through the local GUI. This would make the workflow much more efficient for users who rely on lightweight devices.

Describe the solution you'd like

I would like a feature in the Cellpose GUI to connect to a remote server for computation.

Ideal workflow:

  • Launch the Cellpose GUI locally on a laptop or desktop
  • Connect the GUI to a remote machine or server
  • Run model inference on the remote server
  • Train or fine-tune models on the remote server
  • View results and continue annotation in the local GUI

This would separate the compute-heavy tasks from the user interface, allowing users to annotate locally while using server-side CPU/GPU resources for inference and training.

Describe alternatives you've considered

Current alternatives are not very convenient:

  • Running everything locally, which is often too slow on laptops
  • Logging into a remote server separately and running Cellpose outside the GUI
  • Using remote desktop solutions, which can feel laggy and are less convenient for annotation
  • Manually moving data between local and remote environments

These workarounds are possible, but they make the workflow more complicated and less efficient than having direct remote-server support built into the GUI.

Additional context

This feature would be especially useful for:

  • Users with laptops and limited local compute
  • Labs with shared GPU servers
  • Large datasets that require faster inference/training
  • Annotation workflows where responsiveness in the GUI is important

A client-server mode for the GUI could make Cellpose much more practical in research environments where compute resources are centralized on servers but annotation is done on personal devices.

Activity

  1. mrariden commented on Apr 29, 2026

    @mrariden
    Collaborator

    Thanks for the suggestion. It would be a pretty large lift.

  2. mmaleki92 commented on Jun 20, 2026

    @mmaleki92

    Hi,

    We need this exact same feature for our research group. I actually already have a few of our simulation pipelines running as web-based interfaces on our servers, so I would love to develop this for Cellpose.

    I want to build a lightweight web-based interface (likely using Flask) that can be triggered directly from the command line. My idea is to add it as a CLI flag, so running python -m cellpose --web_server --port 8000 on the server would host the GUI. This way, the server handles the GPU inference, and users can just connect via their local browser to do the annotation and run the models.

    To keep the main Cellpose environment clean, my plan is to make the web framework an optional dependency (e.g., pip install cellpose[web]) so it doesn't add bulk for users who don't need it.

    If there is anything specific I need to know, or if you have any preferences on the web framework before I start developing this and put together a PR, I would be happy to hear from you!

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Metadata

Metadata

Assignees

No one assigned

    Type

    No type

    Projects

    No projects

      Milestone

      No milestone

      Relationships

      None yet

      Development

      No branches or pull requests

      Issue actions