- the detection is not great at the moment (15 globules detected amongst the 50), because:
1. the parameters are probably not optimal
2. the mask used is incomplete
3. some particules are very difficult to detect anyway
- added documentation
- still need some more work:
1 output the radius of detected particles
2. analyse why detection is not optimal
3. handle weights w_r and w_a
4. filter out radial profiles that are not expected
5. improve the computation of the peakness s_p
v1.02
nb: my vscode project is at the moment outside git repos, so the fix to the vscode project is not visible in git. But fwiw, I declared PYTHONPATH variable in a .env file at the root of the workspace.
- saving results as hdf5 allows easy usage or exploration of results using hdfview or hdfcompass, or by any that supports hdf5 files.
- I had to add support for hdf5 file export, which was more complicate than using h5py library (which is not available in fiji's jython). So instead of using h5py, I wrote a generic hdf5 data structure (not restricted to jython), that can be saved using the jython jhdf5 library.
This allows to now detect if the test succeeds or not. Also turned test_preprocessing into a proper unit test. This test is now much cleaner, as preprocessing.py does no longer require ij related code.
As a result, the implementation of estimate_white is getting closer to completion. However, the performance of fiji's gray morphology is very poor, which makes it practically unusable with the default settings provided in the original code (estimateWhiteFluoImageTelemos)
Bug 2623 - Faire un traitement automatique pour les images du projet lipase
- started to write estimate_white, with its unit test
- refactor: split lipase.py into modules
- cleanup : removed unused files and unused jenkins steps
- improved jenkinsfile
Bug 2623 - Faire un traitement automatique pour les images du projet lipase
Guillaume Raffy
Guillaume Raffy