1. petermr
  2. diagramanalyzer

Commits

Author Commit Message Date Builds
petermr
more diagrams added
petermr
removed debugs
petermr
minor tweaks
petermr
added error traps to tree building
petermr
checked for null image
petermr
added missing source
petermr
cleaned phyloTree
petermr
tidying output and added some HOCR stuff
petermr
added missing DiagramTree
petermr
got PhyloTreePixel analysis working
petermr
cleaned pom of redundant dependencies
petermr
removed some of most CPU-intense tests
AndyHowlett
Method DiagramAnalyzer.main() now throws a object of type Exception, so overriding methods can throw anything of that type.
AndyHowlett
Images are now put through the OCR engine correctly regardless of whether they are flipped or not. Switched to SampleModel.getNumBands() from SampleModel.getNumDataElements(); hopefully that is correct. Tidied one piece of code.
AndyHowlett
Changed tests to reflect changes elsewhere. Added a new test; method testMolecularAnalyzerList() now takes a list, which is used by the new test and the old version of method testMolecularAnalyzerList().
AndyHowlett
Added new class OCRManager; its functionality was originally in ChemistryBuilder (should have been part of earlier commit).
AndyHowlett
Added new dependencies to POM.
AndyHowlett
Added OCR to DiagramAnalyzer.
AndyHowlett
Continued work on interpreting diagrams (no text comprehension yet).
AndyHowlett
Added new class PrincipalComponentAnalysis (could maybe move this elsewhere).
AndyHowlett
Improved code for finding starting pixels.
AndyHowlett
Hopefully the logic for shapes not involving cycles is now correct (involved switching from objects of class Iterator to objects of class ListIterator).
AndyHowlett
Code for finding starting pixels now correctly handles a single ring as input (should have been part of earlier commit).
AndyHowlett
Corrected code for handling skipping and adding pixels either side of the skip.
AndyHowlett
Orthogonal branches with a width of one pixel next to the outline of the shape being generated are now correctly handled.
AndyHowlett
When skipping the code now checks that more than one pixel is processed. Formatting change.
AndyHowlett
Failure to find shapes now means the outer shell is used (this needs more work). Corrected code for finding starting pixels and corrected code for iterating onwards from these to avoid lack of convergence and incorrect claims of lack of convergence. Tidied one piece of code and changed a variable name.
AndyHowlett
Added code to cap cut rings when finding polygons (uses Bresenham's line algorithm).
AndyHowlett
Continued work on the shape-finding algorithm (including adding skipping).
AndyHowlett
Any line which doesn't have its midpoint and quarter-points within a polygon but does intersect it is truncated by removing the obscured part.
  1. Prev
  2. Next