PyImageJ working in JupyterLab_summary

[xp1632]

• We already did some research for PyImageJ in Isuues #18 #47 :
• Now we'll test the PyImageJ examples in Jupyterlab with IPython Kernel

---

Our long-term goal:

• [x] understand how PyImageJ works
• [x] PyImageJ Ops run in Jupyter
• [x] PyImageJ custom plugins run in Jupyter
• [x] Python code of calling ImageJ Functions ---> JSON in Visual Programming Project

---

• [ ] Knowledge Graph--> java imageJ to python imageJ part

[xp1632]

• [x] We first check the PyImageJ documentation: https://py.imagej.net/en/latest/api.html in https://github.com/Max-ChenFei/VPE_IP/issues/61

• [x] We then install PyImageJ in our env via Mamba in https://github.com/Max-ChenFei/VPE_IP/issues/62#issuecomment-1904525684

• [x] We go through the jupyter notebook of PyImageJ https://github.com/imagej/pyimagej/blob/main/doc/01-Starting-PyImageJ.ipynb

• [x] Then test the providing PyImageJ Usecase: https://py.imagej.net/en/latest/segmentation-use-cases.html

• [x] We check how to add customized Plugins in PyImageJ

---

• [x] There are also two other whole processes of PyImageJ tutorial in video:
• [x] Video 1 PyImageJ: Integrating ImageJ and Fiji with tools in the Python ecosystem https://www.youtube.com/watch?v=chue-u3RpBM ( 1 year ago)
  • this video has similar content with the second one, we'll only check the installation part
• [x] Video 2 PyImageJ: Integrating ImageJ/Fiji with the Python ecosystem https://www.youtube.com/watch?v=_dL_q0acmYg (1 month ago)

[xp1632]

Useful Informations in Video 2:

---

• PyImageJ has built-in python<->java data type converter as shown in the image

---

---

JImport

• We can import the package for Python in our ExternalImports

• And the import for Java will be inside the code, which is calling jimport method

• Also, the ops method is exactly the same whether we call it in ori ImageJ or PyImageJ

---

• Java <--> Python:

---

import infos about numpy,ndarry -> java convert

• For list, as in the screenshot below,

  • the python list and the converted java list are separated.

• However, in the current version of ImageJ, when we have a ndarray or xarray in Python.

  • then converted it into java
  • when we modify the ndarray, the java object will also change
  • because the java object is linked to the python one

• But it will not happen in another direction, when we java-->python, it's a copy, not linked

• reference from https://www.youtube.com/watch?v=_dL_q0acmYg 26:26

[xp1632]

• Read image via io is also the same, the only difference is that we convert the datatype java<-> python

---

Metadata problem for image,【useful for knowledge graph】:

• the order of metadata could be different in ImageJ image and numpy, xarray
• detailed problem of showing image in wrong metadata order, is shown in video around 40:39 https://www.youtube.com/watch?v=_dL_q0acmYg

[xp1632]

Thoughts on Interactive mode?

• PyImageJ offers headless mode and interactive mode,

  • My original thoughts are we'll only use headless mode
  • but if in some extent, have ImageJ UI means we could use ROIManger
  • No, but having UI would confuse the user where should the actual input output locate, in VP or ImageJ UI

  ---

  • Thus we only use PyImageJ with headless mode

[xp1632]

• There are different categories of PyImageJ, I'll divide them into sub-issues

  • API: https://py.imagej.net/en/latest/api.html#initialization
  • tutorial notebook: https://py.imagej.net/en/latest/notebooks.html
  • usecase: https://py.imagej.net/en/latest/segmentation-use-cases.html