-
Notifications
You must be signed in to change notification settings - Fork 3
Your Model in Fiji
This page describes shortly what you can do to run your own TensorFlow model for data reconstruction in Fiji.
Please be aware that these plugins are currently being updated to match the Python workflow (src, doc).
You will need to export your trained model as a SavedModel. You can find information on how to do that in the Python code documentation
If you trained your model in Keras this GitHub gist will help you to export it as a SavedModel.
You can just execute the Run your network command which will allow you to load your saved model as a zip file and execute it on the open image with an arbitrary "image dimensions <-> tensor dimensions" mapping. In theory, this command could handle all kinds of networks, currently we only tested 2D -> 2D and 3D -> 3D image networks. Write us if you have other ideas and need help executing them.
Take a look at one of our Commands in the CSBDeep Plugin on GitHub. We are working on simplifying the process of writing custom commands even more. Also have a look at the scripts to get an idea how to run the Commands from Python / Jython.