Join BioImage.IO as a community partner

[oeway]

No description provided.

[acherman]

Hi!

We're looking to be added as a community partner!

Our repo has a few more details, as well as the (hopefully correctly formatted) manifest yaml: https://github.com/acherman/fat_checker_work

Broadly, we're training a model to classify lipids in electron micrographs of liver sections. We'd like to be added to your platform to facilitate user analyses with Fiji.

Happy to discuss further with anybody and include collaborators from your group.

• Adam

[oeway]

Hi @acherman Thanks for considering adding your model to BioImage.IO! We are definitely willing to help you with that. However, I have to warn you that the project is still in early stage, so please expect changes in the format and also we are exploring the workflow to accept models from the community. Your early participation will definitely help us shaping how things work.

With that said, if I understand correctly, you basically want to add your fat checker model and notebook (I got to know Etch A Cell project through Martin's poster BTW) to the model zoo as two cards, correct? It's great that you already had some attempts on making the yaml file. I have the following questions:

1. Are you planning to have a series of models for the fat checker? I ask because community partner is meant for groups/software who will constantly produce models. Otherwise if it's just one or two models then we can only take the model but you won't be join as a community partner. (At least this is my interpretation, others please discuss here if you have a different opinion)
2. From the documentation, it seems your model can be used in deepimagej (or fiji?), in that case, the current policy is that you should send a pull request with your model yaml file to the deepimagej models repo. Since that repo is further linked to Bioimage.IO, your model (and notebook) will show up in bioimage.io. For that @esgomezm may help you with the format for deepimagej, and the basic yaml file format is here, but you can use the deepimagej tool to generate this file automatically.
3. It seems you are using google drive to store the models (and I cannot access it without permission BTW), and we would recommend to use Github releases (upload the model weights to the releases page of your repo) or Zenodo to store the model to make sure they are reliably served.

Now is the vacation time, but feel free to join our weekly model zoo meeting (see bioimage-io/bioimage.io#28), suspended until 15th Jan. 2020 (3pm CET).

[constantinpape]

Hi @acherman and thanks for your interest!

I agree with @oeway that whether or not it makes sense for Fat Checker to join as a community partner depends on the scope of what you want to add:

• If you just want to add a few models and notebooks and use deepImageJ (or some other tool already supported by the model zoo community partners), then it probably doesn't make sense to join as a community partner. Of course you can add all your models and notebooks, as @oeway has outlined and all the credit for the model will be assigned to you in the model cards; you can make your models/notebooks easily discoverable and searchable by adding a unique tag, like fat-checker to them in the model yaml.
• If you plan to have a large number of models and your own tool that uses them, then it would make sense to join as a community partner.

Note that you can also start adding your first models and join as a community partner later if the scope of your project widens and you have more models and tool(s).