ivigraph

ivigraph and interactive version of vigranumpy with a labView-ish gui for image processing operator flow charts

Quickstart:

run flow.py and change the "lena" path to an image of your choise. Cick on Flowchart-DockArea (flowchart itself might be hidden) in the GUI and edit the graph

• right click to add new operators
• press "del" to remove connections and nodes
• connect operator to one of the 4 image viewers

Dependencies:

• numpy

• vigranumpy http://hci.iwr.uni-heidelberg.de/vigra/ Python and Boost Python Wrapped C++ Code

  • one currently needs a very new version if vigra .
  • Use the newest version from https://github.com/ukoethe/vigra master

• pyqtgraph http://www.pyqtgraph.org/ pure Python

• termcolor

  • sudo easy_install termcolor

• scipy

Known Bugs:

• ColorChannel Nodes do not work on some machines ( strage vigra c++ signature error)
• Load:
  • one might need change the histogram range a tiny bit to refresh the image
• Sometimes one gets floating point error when deleting edges no idea what is going on
  • (can someone reproduce this error?)

Authors

• Thorsten Beier
• Philipp Hanslovsky

Version

• 0.0.6.11

  • new node gui is intruduced in the first real node (numpy require node as first usecase)

• 0.0.6.10

  • fixed tensor nodes a bit ( some of them work with colorchannel images now)

• 0.0.6.9

  • batch mode image selector combo box is working

• 0.0.6.8

  • viewer will autoselect best mode for image

• 0.0.6.7

  • experimental region features (dummy,does not work so far)
  • added make list , make tupe node

• 0.0.6.6

  • refactored node base classes
  • split operators into seperated files
  • removed some unused imports

• 0.0.6.5

  • fixed minor bug in vigra_machine_learning.py

• 0.0.6.4

  • use MyNode and MyCtrlNode as base for allmost all operators

• 0.0.6.3

  • removed stupid prints
  • added colored printing (needs termcolor)
  • improved base class for ctrl nodes
  • added timing to "vigra" nodes
  • (hot-) fixed a so far unknown loading bug (error was raised but catched)

• 0.0.6.2

  • added a few demos

• 0.0.6.1

  • implemented a "batch-mode" to have a folder with images as input .
  • image selection gui is enabled if input mode is batch mode

• 0.0.6.0

  • implemented a central widget ivigraph

• 0.0.5.9

  • added superpixel visualization
  • added a "demo" folder to repo where nice example flowcharts can be stored

• 0.0.5.8

  • removed unused buttons: (flowchart,reloadLibs,Norm)
  • added experimental gui for image selection (only GUI, no functionality so far)
  • refactored Viewer a bit
  • added label alpha slider

• 0.0.5.7

  • added slicSuperpixels (and RegionToEdges for visualization)

• 0.0.5.6

  • added Random Forest classifier nodes (including auxiliary nodes)

• 0.0.5.5

  • added numpy whereNotNode
  • started to make it easy to implement a generalize numpy.where

• 0.0.5.4

  • added numpy require node

• 0.0.5.3

  • readded the functionality to switch between rgb,lab,and 1-channel
  • removed unused ui templates from viewNodes

• 0.0.5.2

  • proof of concept for machine learning with random forest

• 0.0.5.1

  • labeling uses multiple ImageItems in a layerd fashion

• 0.0.5.0

  • INITAL HACKY VERSION OF SAVE AND LOAD IS WORKING

• 0.0.4.1

  • added channel stacker with flexible number of input terminal

• 0.0.4.0

  • changed main layout to a "dockarea"

• 0.0.3.5

  • improved and fixed clear labels

• 0.0.3.4

  • improved viewer gui

• 0.0.3.3

  • extended imageview functions

• 0.0.3.2

  • labeling is allmost finished for rgb color images

• 0.0.3.1

  • fixed bug :brush size is now changeable

• 0.0.3.0

  • proof of concept for labeling is done

• 0.0.2.0

  • code works on differnt machines

• 0.0.1.0

  • initial version