search

vanvalenlab / deepcell-tf

Deep Learning Library for Single Cell Analysis
https://deepcell.readthedocs.io
Other
412 stars 94 forks source link

Docker image to integrate into a pipeline #372

Closed ngreenwald closed 4 years ago

ngreenwald commented 4 years ago

Is your feature request related to a problem? Please describe. A couple of groups have reached out to me with interest in using the multiplexed model as part of a larger image processing pipeline, as part of something like QuPath or Nextflow. I think the current setup of a jupyter notebook-based docker image is not super conducive to that.

Describe the solution you'd like I know that many of these pipelines can accept appropriately configured docker images that serve as one piece. How much work would it be to create a deepcell-tf image that would be compatible with some of these microscopy analysis pipelines? It wouldn't have to support training in any capacity, just inference of an existing application object.

willgraf commented 4 years ago

It does look like NextFlow supports using docker containers in a pipeline.

ngreenwald commented 4 years ago

That's great. Looks like maybe all that's needed is copying the relevant code from the example applications notebook into a python script that they call, specifying with our container? And giving them a dockerfile that doesn't start up a jupyter notebook?

willgraf commented 4 years ago

The deepcell-applications repository has been created for this.

Currently, it just has an example script using the deepcell Applications, but eventually, all of the applications will be migrated there.