2016-01-29 - KNIME Image Processing hackathon

From Monday, January 11, 2016 through Friday, January 22, 2016, Christian Dietz of the University of Konstanz hosted ~30 developers at the KNIME Konstanz Regional Office for a hackathon to develop the KNIME Image Processing extensions along with the underlying ImageJ Ops library, as well as other supporting technologies such as ImgLib2 and ClearVolume.

Timeline

The following chart illustrates when everyone was present:

Hackathon progress

Mark Hiner

Curtis Rueden

  • Research
    • Investigated improvements to Ops generic parameter matching. The problem is complex, but a solution is in the works.
    • Explored possible Ops matching performance improvements (fast-op-index)
    • Improved Jenkins jobs to build SNAPSHOT javadoc for all core SciJava components (scijava, scifio, imglib2 and imagej). as requested by Stefan Helfrich (1, 2, 3)
  • Reviewed and merged PRs (1, 2, 3, 4, 5, 6, 7, 8, 9)

Robert Haase

  • Implemented continuous (Hyper-)Ellipsoids, Rectangles and Binary operators for imglib2-roi.
  • Developed prototype for “Data Explorer” window for ImageJ/Fiji
    • holding various data structures like images, labelings, measurements (tables) and plots. The structure is persistent, so closing windows does not delete the items from the tree. They can be reopened.
    • 2D/3D/nD? data is handled equally, so selecting/deleting labels for/from further analysis according to their properties is possible more or less independent from the labels dimensionality.
    • Measurements on labels is based on imagej ops()

Patrick Winter

  • org.knime.cluster the abstract implementation of the cluster integration in KNIME is now open source, enabling the community to easily implement their own cluster connector
  • Wrapped metanodes (aka subnodes) can now be executed in the cluster and also support split input data (which was not the case with the previous implementation of metanodes)
  • Implementation of cluster connectors got simpler since we removed the differentiation of single jobs and chunked jobs with a size of 1
  • Many bugfixes in the old implementation, as well as fixes in org.knime.core (mostly concerning subnodes)

Ulrik Günther

Tobias Pietzsch

Stefan Helfrich

Technical discussions

Future directions