5.2. Log of progress 2019

  • Author: Felipe Orihuela-Espina
  • Created: December 17, 2018
  • Revised: January 21, 2018
  • Copyright (c) 2018-9 INAOE

Please note that advances indicated at a particular date, may actually refer to advances in the previous days/weeks.

5.2.1. Advances 3-Nov-2019

  • Version: v0.3
  • Responsible: FOE

Summary of changes:

The project has suffered a 2 months period of inactivity due to other commitments. Activity is now being retaken.

  • Segmentation algorithm progress;
    • Detection of the background mask now working properly.
    • Advanced in the assigment of tissue pixels to retinal layers.
  • GUI
    • Detection of console/terminal or gui mode for showing a progress bar during segmentation annotated as a pending task. FOE have made some progress here, but this is not yet working.

Documentation

  • Class octant.data.RetinalLayers Added some comments about sublayers.

. _secLogAdvances20190819:

5.2.2. Advances 19-Aug-2019

  • Version: v0.3
  • Responsible: FOE

Summary of changes:

  • Segmentation algorithm progress;
    • Started updating the detection of the background mask.
    • The output of the anisotropic diffusion filter has been normalized and discretized so that the values can be related to retinal layers, but this is not yet clear. I’m still looking at the histogram.
    • The anisotropic diffusion filter is now a static function for greater flexibility.
  • GUI:
    • With the new graphical progress bar in method execute() of class octant.op.OpScanSegment, matplotlib backend was no longer automatically launching inline in the IPython console. Now it requires explicit call to:
      • %matplotlib to activate the backend
      • %matplotlib inline to make it work interactively in the IPython console.
    • I have tried to automatically detect whether the program is being executed from the terminal or from the GUI using os.isatty and sys.stdin.isatty but none can really make the distinction. Thus, by now I have “disconnected” the progress bar, while I work, but the code is there to be activated whenever is needed.

Documentation

  • Log of progress split by year for easier management. Collaterally it also improves the indexing of the main table of contents, as it moves one level down the weekly progress sections.

5.2.3. Advances 12-Aug-2019

  • Version: v0.3
  • Responsible: FOE

Summary of changes:

  • The perfilometer axis now correctly resizes with main scan axis.
  • Integration of the diffusion filter in the main GUI, that is, removed intermediate figure plotting, debugging code, etc. Also added a progress bar.

Documentation

  • Documentation update. Sphynx documentation recompiled.
  • Updating of the versioning, planning and toDo.
    • Reorganization of the toDo list to act also for monitoring task progress, kinda ‘tiny’ software manager tool. Not the best option, but the options for software management that I know are all proprietary software. So it will do by now.
    • Reassignment of tasks to version as appropriate.

Bug Fixing

  • DocumentWindow.refresh: Perfilometer was plot “upside down”. It now
    renders adequately.

5.2.4. Advances 8-Aug-2019

  • Version: v0.3
  • Responsible: FOE

Summary of changes:

  • Worked on making the perfilometer axis to resize with main scan axis. Although, it does it transiently, but it flickers and goes back to a different size. So not fully working yet.

5.2.5. Advances 8-Jul-2019

  • Version: v0.3
  • Responsible: FOE

Summary of changes:

  • Automatic segmentation; diffusion filtering. Finally working! In the end there was not a single but several errors, including;
    • Inadequate parameterization for human retina. [WangRK2005] Parameters were for the porcine trachea.
    • Regularization Gaussian mask was increased from 3x3 to 7x7 and Updated every integration.
    • Update of the image on each iteration was not being done properly.
    • The morphological filter has been recovered but with a twist. It is now applied on every iteration of the diffusion filter.

5.2.6. Advances 4-Jul-2019

  • Version: v0.3
  • Responsible: FOE

Summary of changes:

NOTE: Progress this week was hindered due to problems with FOE’s laptop.

  • Automatic segmentation; diffusion filtering. After several attempts to check for errors in the programming without success in preventing the diluting of the gradient, an alternative hypothesis is that since the parameterization in [WangRK2005] is for the porcine trachea, these parameter values may not be good for the human retina. I have started to look for parameterization in human retina, but so far no joy.

5.2.7. Advances 24-Jun-2019

  • Version: v0.3
  • Responsible: FOE

Summary of changes:

  • Attempted to improve the performance of the segmentation algorithm by substituting the initial noise removal step from the current morphological closing of opening to the more specialized non-linear anisotropic diffusion filter as first described by [P. Perona and J. Malik, IEEE Trans. Pattern Anal. Mach. Intel. 12, 629 (1990)] and later reported for OCT by [Wang, RK (2005) Proc. SPIE 5690:380-385].
    • This is still NOT working. Unfortunately, the gradient seem to be diluting to quickly, and thus output image is distorted in scale. After, some thinking and testing I think the error can be in our gray images being scaled [0,1] but [WangRK2005] perhaps using a [0,255] scale. Another, potential explanation is the iterative nature of the Gaussian mask (see [Salinas and Cabrera-Fernandez (2007) TMI, 26(6):761-771].

5.2.8. Advances 18-Jun-2019

  • Version: v0.3
  • Responsible: FOE

Summary of changes:

  • Added codification of fluid as an additional segmentation layer in RetinalLayers for pathological cases.
  • Added colormap of retinal layers to settings in configuration file as settings retinallayerscolormap. The chosen default colormap is according to [ChiuSJ2015_BOE] with the minor unfolding of the GCL-ILM layers and the consideration of choroid.
  • Documentation
    • Prepared report on the flattening algorithm and uploaded to OSF.io in the Results component.
    • Uploaded several new references to OSF.io
    • Fixed link to OCTant in GitHub from OSF.io. It was still pointing to old oct-tools.
    • OSF.io main project component renamed to also include OCTant in the title.
    • Initial check on Duke dataset [ChiuSJ2015_BOE]. Exploration thus far is being made in MATLAB (as this is the original format in which the dataset is released). Advances so far are being reported in file DukeOCTDataset_2015_BOE_Chiu_README.txt, but further work is still necessary to fully decode how the dataset is encoded.

5.2.9. Advances 10-Jun-2019

  • Version: v0.3
  • Responsible: FOE

Summary of changes:

  • First attempts to improve the segmentation algorithm in v0.3. Created a minimal sandbox for fast testing. In this process;
    • Added import of individual classes to make the app available as a package.
    • We learn that python doesn’t make on demand deep copies of objects but that these have to be made explicitly. No big deal (although a bit unexpected since it does on demand deep copies of the built-in objects) but it puts us in the cross-road to whether we need to provide all classes with a copy constructor or a copyobject class method. By now the sandbox will continue using the package copy and method deepcopy.

5.2.10. Advances 1-Jun-2019

  • Version: v0.3
  • Responsible: FOE

Summary of changes:

  • Commit executed : “Operations model based on OCTvolume stable”
  • Added several new properties (ratio, projThresh and showMatches) to Stitcher in panorama. It now keeps track of whether the operands were switched, and also remembers parameterization.
  • New method readFile in octant.data.Document preparation for persistence. Still naive though.
  • Class octant.op.OpScanStitch underwent several changhes;
    • Added property switchedOperands to flag whether switching the order of the operands was needed during stitching.
    • Added property sparedCols to mark the size of the “black” spared region.
    • Added method applyStitch to repeat a known sticthing to operands. This can be used to apply the same stitch to a different set of scans. In practical terms, it can be used to apply the same panoramic stitching to segmentation scans after it has been precalculated to anatomical scans.
  • Class octant.op.OpScanFlatten underwent several changhes;
    • New read only property deformation map to store the deformation map associated with the flattening operation.
    • New method applyOperation to repeat a known flatenning operation to operands. This can be used to apply the same flattening to a different set of scans. In practical terms, it can be used to apply the same flatenning to segmentation scans after it has been precalculated to anatomical scans.
  • Class octant.app.DocumentWindow underwent several changhes;
    • Operation to stitch now also stitches the segmentation.
    • Operation to flatten now also flattens the segmentation.
    • New method _readOCTantFile for reading OCTant files
    • Method _openDocument deprecated in favour of method _readOCTantFile to avoid confusion with new method openDocument.
    • Method openFile is now static and does not modify the current document. Instead, new method openDocument wraps openFile and absorbs the non static operation, modifying the current document.
    • Method _importImageFile is now static.
    • Method _getSemiTransparentColormap is now static.
    • Method _getFilename is now deprecated.

Bug fixing

  • OpScanMeasureLayerThickness
    • Indexing of window was being made from from rows instead of columns.
  • Class octant.app.DocumentWindow
    • Reading second document during stitching was also modifying the current document segmentation because of side effect from openFile not being static.
  • Class octant.app.ToolsDock
    • Upon enabling the segmentation edit, a new dummy segmentation was always being created even if one already existed.

5.2.11. Advances 19-May-2019

  • Version: v0.3
  • Responsible: FOE

Summary of changes:

Currently implementing the collateral stitching of the segmentations, but yet unfinished.

New features

  • class:octant.data.Document
    • Properties study and segmentation are now initialized to OCTvolume and OCTvolumeSegmentation respectively.
  • DocumentWindow
    • Importing scans from an external file format now also initializes the document segmentation property with an empty segmentation volume of the same size that the imported OCTvolume.
  • Added new property homographyMatrix in octant.op.OpScanStitch. This permits transmitting the stitching to the segmentation scans.
    • This has required making this information accesible in panorama. Hence, I have also added an analogous new property homographyMatrix to octant.util.Stitcher.

Bug fixing

  • DocumentWindow
    • Call to segment operation was assigning operand of octant.op.OpScanSegment to the wrong object.
    • Reference to the octant.data.RetinalLayers class constructor in the refresh() method was not indicating the subpackage.
    • Retrieval of current segmentation scan the refresh method was incorrectly pointing to the wrapping volume.
    • Remaining references to old attribute _toolsWindow updated to _toolsDock.
    • Method measureThickness was setting the segmentation volume instead of the segmentation scan as the operand for operation octant.op.OpScanMeasureLayerThickness
  • class:octant.op.OpScanSegment
    • Method execute() was not testing for number of operands correctly.
  • class:octant.data.OCTvolumeSegmentation
    • The flag for testing all inputs in method addScanSegmentation() to be of type OCTscans was not being correctly initialized.
  • class:octant.data.Document
    • Segmentation property setter was incorrectly setting property study.
    • Segmentation property setter was attempting to assert the number of scans against the study reference using shape instead of len.
    • Methods getCurrentScan’ and `getCurrentScanSegmentation were not checking for empty scan lists.

5.2.12. Advances 13-May-2019

  • Version: v0.3
  • Responsible: FOE

Summary of changes:

Miscellaneous

  • Data model with document based on octant.data.OCTvolume is now considered stable. * Commit executed : “Data model based on OCTvolume stable”

Bug fixing

  • Class app.ToolsDock no longer import app.DocumentWindow breaking the circular import.
  • Class app.UtilitiesDock no longer import app.DocumentWindow breaking the circular import.

Documentation

  • Recompiled documentation.
    • Bug fixed. Sys.paths in Sphynx Conf.py for package app was unable to find the path ….srcapp because a is a escape character. This was causing that the documentation of some of the classes in this package were not being built correctly.
    • Fixed some minor Sphinx related typos/mistakes in several files, including this one, utilsegmentationUtils.py, intro.rst
    • Added inheritance diagrams to classes in package app

5.2.13. Advances 6-May-2019

  • Version: v0.3
  • Responsible: FOE

Summary of changes:

New features of the app:

  • Operands passed to stitching can now be given in an arbitrary order. Before, the first image always had to be the right most in the mosaic.

Bug fixing

  • The “bug” in panorama apparently related to an issue with images depth, and that happened to open Pandora’s box with changes coming from the latest OpenCV v4.0, and that we have been dealing with in the last several weeks has finally been solved! Last week, we were already in the right path by switching to ORB features, and the updating of the syntax to OpernCV v4.0 but depth conversion to uint8 although mathematically responsive (it yielded no error) but was not giving correct results and thus the ORB feature detector failed to detect any features. Finally, This week, we succeded in getting the depth conversion right (anecdotically we almost got it right last week but we were scaling by 255 after downcasting instead of before downcasting). So in summary, the bug a mixture of a real issue with images depth (necessitating scaling followed by downcasting), new syntax in OpenCV v4.0, the disappearing of SIFT feature detector as a free option in OpenCV (requiring adaptation to a different feature detector), and finally some parameter tuning (e.g. new value for parameters on the feature detection and matching given the new feature detector algorithm). The latest parameters are:
    • nfeatures=100000 (this degrades speed a bit, so may be worth adjusting a bit more to the smallest number we can).
    • number pyramid levels = 16
    • fastThreshold=10
    • scoreType=cv2.ORB_FAST_SCORE
    • ratio for keypoint matching = 0.9

5.2.14. Advances 29-Abr-2019

  • Version: v0.3
  • Responsible: FOE

Summary of changes:

Bug fixing (or not)

  • A new bug appear in module panorama with the upgrade to OpenCV v4.0. At first sight, it looks like a simple problem of panorama not being to handle depth of 64 bits in the images, but then that’s only the tip of the iceberg. SIFT feature descriptors are no longer free, and the syntax to create the feature detector has changed with the new version of OpenCV. We have tried several things, but yet without full success;
    • Added support to panorama for OpenCV v4.0.
    • Changed the feature detector from SIFT to ORB.
    • Depth of images has been brought down to uint8 from float64. That permits running the mosaicing without error, but the descriptor then produces no features. Tested on a NON retinal image, ORB seems to be working fine. So perhaps ORB is not good for retinal images.
    • Changed the feature detector to BRISK. Same results as with ORB; no descriptors on the retinal image.
    • Attempted image normalization, but then panorama crashes again.

5.2.15. Advances 9-Abr-2019

  • Version: v0.3
  • Responsible: FOE

Summary of changes:

New features of the data model:

  • New methods getCurrentScanSegmentation and setCurrentScanSegmentation in octant.data.Document

New features of the app: * Several changes to octant.app.DocumentWindow

  • New method _openDocument to read OCTant documents. Since serialization is not yet ready, by now it yields a warning and returns an empty document.
  • Method importAmiraFile renamed openFile, as it was not actually neither assuming that it was an Amira file, nor that it was an importing operation. Further, it now distinguishes OCTant file extension to bifurcate execution to call either importFile when the file is in an external format, or _openDocument when it is in OCTant document format.
  • Method _getImageFilename renamed _getFilename. Also, Sphinx styled comments have been added.
  • Bug fixed. Stitching was still calling “old” method openDocument. This is a double bug; first, the method name should have been importAmiraFile (now openFile), and second, because it assumes that the 2nd document has to be imported from an external format, rather than read from my format. Of course this is fine while we develop the document serialization, but nonetheless, but should anticipate. Now it calls either _openDocument or _importImageFile as appropriate.
  • Method importAmiraFile renamed importFile in octant.app.ToolsDock

Documentation

  • Added pending feature for v0.4: Allow selection of scan for stitching. Currently stitching is made against default selected scan.
  • Annotated in toDo list detected bug in panorama.py regarding unsupported color depth.

Bug fixing

  • References to octant.data.OCTscan in octant.op.OpScanFlatten updated. updated. * References to octant.data.OCTscan in octant.op.OpSegmentationEdit
  • Method editSegmentation in octant.app.ToolsDock was still using “old” property documentWindow. In now calls method parent().
  • Constructor in octant.data.OCTscanSegmentation was still making reference to IOT_OCTScan.
  • Call update from OpEditSegmentation._BACKGROUND to OpSegmentationEdit._BACKGROUND in method _generateDummySegmentation in octant.data.OCTscanSegmentation
  • Subpackage octant.data was not exporting octant.data.OCTscanSegmentation in __init__.py
  • Call to study.addScanSegmentations updated from study.addScanSegmentation in method octant.data.Document.segmentation(). Also, parameter passed is now correct.

5.2.16. Advances 1-Abr-2019

  • Version: v0.3
  • Responsible: FOE

Summary of changes:

New features of the data model:

  • Property currentScan in octant.data.Document replaced by methods getCurrentScan and setCurrentScan.

New features of the app:

  • Button bOpenImage in octant.app.ToolsDock renamed to bImportImage and relabelled to Import image
  • Main window now opens maximized
  • Utilities dock moved to bottom and left bottom corner conflict resolved.
  • Started class octant.app.ScansCarousel for visualization of OCT scans in a octant.data.OCTvolume and selection of current scan.
    • Bug pending. Although loading of scans is correct but rendering of the thumbnails is not.
  • Added tab to UtilitiesDock to hold the octant.app.ScansCarousel
  • octant.app.DcoumentWindow: Importing file also updates scans carousel in utils dock.

Documentation

  • Added log to module segmentationUtils
  • Fixed comments of property docwindow in octant.app.OCTantApp which were referring to property settings.

Bug fixing

  • Attribute __version__ now imports correctly from octant.data.Document and octant.data.OCTvolume .
  • octant.data.OCTvolume now correctly imports deprecation.
  • octant.data.OCTvolume flagAllOCTScans in method addScans is now correctly returned in all cases.
  • octant.app.DocumentWindow Importing image from common image formats in _importImageFile now ensure that the third dimension corresponds to scans and not to RGB filters.

5.2.17. Advances 25-Mar-2019

  • Version: v0.3
  • Responsible: FOE

Summary of changes:

New features of the data model:

  • Class octant.data.OCTvolume has seen several changes;
    • Changed calls to isinstance for calls to type.
    • Added method getClassName.
    • Added method addScans.
    • Added method getNScans.
    • Deprecated method addScan.
  • New class octant.data.OCTvolumeSegmentation.
  • Class octant.data.Document has seen several changes;
    • Added properties docsettings.
    • Started migration to OCTvolume based document.
    • Added new docsetting .selectedScan
    • Added read only property currentScan
    • Added method pickScan.

New features of the app:

  • Class octant.app.DocumentWindow has seen several changes;
    • Method _openImageFile renamed to _importImageFile.
    • Also, it now returns and OCTvolume rather than a set of scans or an isolated of scan.
    • Method importAmiraFile updated to call _importImageFile.

5.2.18. Advances 18-Mar-2019

  • Version: v0.3
  • Responsible: FOE

Summary of changes:

Landmarks reached for v0.3:

  • Create support for application-wide Settings

New features:

  • Class octant.data.Settings can now read and write files. JSON file format has been chosen for settings files.

    • Note that new dependencies re and json for reading/writing JSON files are both built-in features of python, and hence do not need to be declared in installer.cfg

  • Added JSONminify module to package octant:util

  • Class octant.app.OCTandApp new properties appsettings() and appsettingsfile()

  • New file resources/OCTantApp.config for persistency of application settings. Currently, only workingdirectory property has been set.

  • Launching BAT file renamed to OCTantApp.bat and updated.

  • installer.cfg updated for new BAT, link to new icons and new config file.

  • Menu previously in tools window, has now been moved to application main window in octant.app.DocumentWindow

  • Method openDocument() in class octant.app.DocumentWindow renamed importAmiraFile().

  • New property parentapp in class :class:`octant.app.DocumentWindow() connecting with the main application object.

  • Class app.OCTantApp now becomes a QApplication (previously we had 2 separated objects; one for the QApplication and another just for “holding” the main window.) and underwent several changes:

    • Added properties appsettings and appsettingsfile.
    • docWindow attribute converted to docwindow property.
    • Cleaner exit with call to deleteLater
    • Removed method show. Now the document window show is called accesing the docwindow property.

5.2.19. Advances 12-Mar-2019

  • Version: v0.3
  • Responsible: FOE

Summary of changes:

New features:

  • New class octant.data.Settings for handling document settings.

    • The dynamic struct aspect appears to be working fine.
    • read() method advanced but unfinished.

  • Given initial considerations to web-based implementation with Django (thanks PHW for the tips!)

Bug fixing:

  • Fixed: Attempting to open a new scan when one is already open, will launch the opening dialog, but this will be freezed. The opening dialog does no longer freezes and the new scan is loaded correctly.

  • Attended bug regarding stitching more than 2 images as well as panorama attribute error;

    • panorama.py, line 67: AttributeError: module ‘cv2.cv2’ has no attribute ‘FeatureDetector_create’

    It turns out, both issues were related. The source of the problem was that since OpenCV version 3.0 algorithms that are either patented or in experimental development (which is the case of FeatureDetector_create) were not included/installed by default with package opencv-python and instead required package opencv-contrib-python (see: https://www.pyimagesearch.com/2015/07/16/where-did-sift-and-surf-go-in-opencv-3/ ). Further, packages opencv-contrib-python and opencv-python are incompatible hence requiring uninstalling package opencv-python before uninstalling opencv-contrib-python containing the contrib modules. Finally, to make things worst, the:

    pip uninstall opencv-python

    in my case left a corrupted package leaving pip itself in a corrupt state (pip list will crash), and without any error message indicating the offending corrupt package causing the issue. It turns out, that although the latest version of pip already resolves this issue, “people might still be experiencing this issue because of directories that were corrupted before (or getting corrupted for a completely different reason)” (see https://github.com/pypa/pip/issues/6194 ). As indicated in this reference, finding the corrupt package has to be done “by hand”. This requires going to C:\ProgramData\Anaconda3\lib\site-packages\ and looking for packages folders with a leading ‘-‘ in their names, and manually removed them. After this, pip comes back to life and opencv-contrib-python can be now installed:

    pip install opencv-contrib-python

    After successful installation of opencv-contrib-python both of the above issues were resolved.

    Please note that this refer the bug when trying to stitch the 3rd image in pairs of 2. The fixing does not attend the desired feature for stitching several images at once (as this is NOT a bug but a limitation of panorama as indicated in the summary section Adrian Rosebrock’s (creator of python’s panorama code) article:

Documentation

  • Reorganized toDo.rst in sections
  • Added new pending features e.g. document class defaulting to volume and need for scan navigation panel.

5.2.20. Advances 5-Mar-2019

  • Version: v0.3
  • Responsible: FOE

Summary of changes:

  • OCT-tools officially rebranded as OCTant. GitHub repository name updated, and application rebranded as OCTantApp.

  • The versioning of the GUI shell and of the API are now separated. For simplicity however both the API and App have been assigned v0.3, but they will evolve separatedly from here onwards.

  • New logo and icon designed.

  • New package architecture in development. The previous prefix IOT in class names is now abandoned as classes are packaged. The new package structure now clearly separates the API from the app, and within they API, the data model classes are further separated from the operational classes. The folder structure is left as follows:

    src/
 |- app - The application. This is just a shell over the API.
 |- octant - The API
  |- data - Classes of the data model
  |- op - Operational classes. These are the classes that provide functionality to the package.
  |- util - A misceallaneous of additional functions and external dependencies

  • All classes have been moved to their corresponding folder. The classes corresponding to operations are suggested to follow a naming convention indicating the main operand type before the operation name.

  • All classes have now been migrated to the new architecture pending testing:

    Old class name New class name
    IOT_Document octant.data.Document
    IOT_OCTscan octant.data.OCTscan
    IOT_OCTvolume octant.data.OCTvolume
    IOT_OCTscanSegmentation octant.data.OCTscanSegmentation
    IOT_RetinalLayers octant.data.RetinalLayers
    IOT_Operation octant.op.Operation
    IOT_OperationFlattening octant.op.OpScanFlatten
    IOT_OperationMeasureLayerThickness octant.op.OpScanMeasureLayerThickness
    IOT_OperationSegmentation octant.op.OpScanSegment
    IOT_OperationStitch octant.op.OpScanStitch
    IOT_OperationPerfilometer octant.op.OpScanPerfilometer
    IOT_OperationBrush octant.op.OpSegmentationBrush
    IOT_OperationEditSegmentation octant.op.OpSegmentationEdit
    OCTToolsApp app.OCTantApp
    IOT_GUI_DocumentWindow app.IOT_GUI_DocumentWindow
    IOT_GUI_ToolsWindow app.ToolsDock
    IOT_GUI_UtilitiesDock app.UtilitiesDock
    IOT_GUI_EditSegmentationTools app.OpSegmentationEditToolsPanel
    IOT_GUI_BrushParameterSettings app.SettingsGUIOpSegmentationBrush
    IOT_GUI_MeasureThicknessParameterSettings app.SettingsGUIOpScanMeasurementThickness
    IOT_GUI_PerfilometerParameterSettings app.SettingsGUIOpPerfilometer

  • The class octant.op.Operation now provides support for parameters.

  • The previously deprecated “original” methods for calling the operation have now been fully removed. The use of method execute() is now compulsory.

  • Several calls to isinstance() have been changed by calls to type().

  • Tools window is now a child dock of DocumentWindow which is left as the only QMainWindow of the app.

  • Documentation updates:

    • Updated project README.md
    • Updated intro.rst
    • Updated toDo.rst
    • Updated conf.py
    • Updated installation.rst
    • Updated technical.rst

  • Commit executed : “OCTant Rebranding and repackaging”

5.2.21. Advances 25-Feb-2019

  • Version: v0.3
  • Responsible: FOE

Summary of changes:

  • Reading about subpackaging.

  • Planning separation of foundational classes to a separate project because, in giving priority to Rodrigo’s request, OCT-tools has naturally departed from them and hence the project is not using them. Consequently, the following folders and files have been removed from GitHub repository:

    • docs/EclipseModelling/
    • docs/source/sciMethFileFormatSpec

    Older versions of these files can still of course be found in previous commit history. Documentation of the new architecture is needed. New Eclipse documentation will be now move to OSF project site.

  • List of pending tasks has been moved to To Do.

  • Version: v0.2beta

  • Responsible: FOE

Summary of changes:

  • Refreshed Sphinx documentation.

  • The project will no longer be using Git LFS to avoid incurring in charges. As a result, Git LFS is being uninstalled. Instead, as from now, installers will be hosted in the OSF project site.

  • For a more correct use of GitHub, the following directories have been declared in .gitignore to be no longer tracked, and hence are neither committed/pushed:

    • __pycache__/
    • obsoleteOrTestingCode/
    • docs/build/

  • Version v0.2 is now considered stable and fully released.

5.2.22. Advances 19-Feb-2019

  • Version: v0.2beta
  • Responsible: FOE

Summary of changes:

  • Version v0.2 beta has been committed and pushed to GitHub
  • Installer generation tested on additional computer with a previous version of conda. It failed to compile because of an issue with packages certificates. A simple certificate update did not fixed the problem. A full update of miniconda might be needed.
  • Upgraded version of python set in the installer.cfg from v3.6.5 to v3.7.1. This additionally demanded new wheeled versions of packages:
    • wrapt 1.11.1 - Our previous version was compiled for python v3.6.5
    • imutils 0.5.1 - Updated to v0.5.2
  • New installer for python v3.7.1 ready and shared with Rodrigo.

5.2.23. Advances 13-Feb-2019

  • Version: v0.2alpha
  • Responsible: FOE

Summary of changes:

  • Transitioned to package “deprecation” (from “deprecated”). This requires finer control of package version. Modified classes are:

    • class:src:OCTToolsApp
    • class:src:IOT_Document
    • class:src:IOT_GUI_DocumentWindow
    • class:src:IOT_OCTvolume
    • class:src:IOT_OperationBrush
    • class:src:IOT_OperationFlattening
    • class:src:IOT_OperationMesureLayerThickness
    • class:src:IOT_OperationPerfilometer
    • class:src:IOT_OperationSegmentation
    • class:src:IOT_OperationStitch

  • Version control is now in version.py (instead of setup.py), and loaded in __init__.py

  • Comments in setup.py and __init__.py updated to Sphinx style

  • Opening message now informs of version.

  • First version of the installer is FINALLY compiling. Installer is now being tested.

  • We are now in v0.2beta, but release to GitHub is pending

5.2.24. Advances 5-Feb-2019

  • Version: v0.2alpha
  • Responsible: FOE

Summary of changes:

  • New attempts to generate the installer; now under Anaconda, have failed. The library “deprecated” continues to give problems for compiling in either versions 1.2.3 and 1.2.4.
  • We’re now seeking alternatives:
    • Package deprecation 2.0.6
    • A shortcut by now may be to defer all deprecations to the next version.
  • Deprecation in general seems to be a more general problem in python; https://www.python.org/dev/peps/pep-0004/ which might partially explain the difficulties in compiling the deprecated module.

5.2.25. Advances 27-Jan-2019

  • Version: v0.2
  • Responsible: FOE

Summary of changes:

  • Version v0.2 alpha released and commited to GitHub.
  • Installer.cfg updated for new package versions.
    • wrapt 1.10.11 -> wrapt 1.11.1
    • deprecated 1.2.3 -> deprecated 1.2.4
  • Generation of installer in progress. The library “deprecated” which works well when interpreted is giving some troubles during compilation. This has been tested in two Windows 10 machines (FOE_INAOE laptop and CHOLULA desktop). Currently, investigating a solution.

5.2.26. Advances 20-Jan-2019

  • Version: v0.2
  • Responsible: FOE

Summary of changes:

  • New class src.IOT_GUI_BrushParameterSettings to support GUI control of the brush operation parameters.
  • Modified classes src.IOT_GUI_DocumentWindow with new method brush to support mouse controlled brush operation.
  • Modified class src.IOT_GUI_EditSegmentationTools: Added button for executing Brush operation.
  • Modified class src.IOT_GUI_ToolsWindow: Added tab in parameter settings panel for hold src.IOT_GUI_BrushParameterSettings.
  • New method setOperand in src.IOT_Operation. The direct benefit is a faster response of the brush. But it is easy to foresee additional uses.

Bug fixing:

  • Major debugging of src.IOT_OperationBrush.
  • Bug fixed in src.IOT_OperationBrush: Assigment of property classMap in property setter was being “assigned” to cm.

Bug detected:

  • Attempting to open a new scan when one is already open, will launch the opening dialog, but this will be freezed.

5.2.27. Advances 17-Jan-2019

  • Version: v0.2
  • Responsible: FOE

Summary of changes:

  • New class src.IOT_OperationBrush defined for supporting the manual modification of the segmentation. It will act as a painting brush. Behaviour ready but pending testing and incorporation of controls to the GUI.

    NOTE; This class is using an algorithm that is different from the one proposed by Arlem in MATLAB.

  • Bugs corrected in src.IOT_OperationEditSegmentation whereby the output of the method was not returning the OCT scan. Although I have not experienced further errors, but I suspect that the fixing is still not fully correct. Some further testing is necessary.

  • Bug corrected in the generation of the dummy segmentation where no background scan was associated to object src.IOT_OCTscanSegmentation even when the src.IOT_OCTscan was present.