Analysis in ImageJ/Fiji using TensorFlow models
Some experimentation with creating ImageJ plugins that use TensorFlow image models.
For example, the one plugin right now pacakges the TensorFlow image recognition tutorial, in particular its Java version into a command plugin to label an opened image.
git clone https://github.com/asimshankar/imagej-tensorflow
cd imagej-tensorflow
mvn compile exec:java
This requires Maven. Typically brew install maven
on OS X, apt-get install maven
on Ubuntu, or detailed
instructions otherwise.
The code here is a proof of concept. It has some silly inefficiencies that would be dealt with real use. For example, the TensorFlow model is loaded on each invocation of the command instead of being loaded once and cached for repeated use.