comvi

Comparative Visualization of Molecular Surfaces using Similarity-based Clustering

Megamol-specific terms/keywords:

• *.mmprj: MegaMol Project files These xml-based files can be invoked (with specific parameters given by the Configurator) with the megamol-console.

• Configurator: C#/Mono Project files creator, with visual drag and drop interface for project generation. MM-Creator creates a xml-file (the *.mmprj files from above) which the megamol can understand.

• StateFile: generated by the configurator upon first startup, holds all kinds of information about the mm-binary. is nessecary if you want to configure your project with the interface

Dependencies:

• python:
  • everything was written in python 3.6
  • not tested in python<3.6 (or 2.7 for that matter)
  • dependencies:
    • numpy
    • tensorflow
    • sklearn
    • skimage
    • scipy
    • cv2
    • matplotlib for 2d plot-viz
• comvi C++:
  • not sure
  • probably only vanilla c++11/14 (?) apart from:
    • glfw
    • glew
    • glad
    • glm
    • std_image
    • stdlib
    • KHR
    • opengl

Folder Structure:

• papers:

  • contains all relevant papers for:
    • clustering
    • viz
    • feature extraction etc.

• comvi Ausarbeitung:

  • contains .tex sourcecode for the comvi paper

• programming:

  • contains:
    • PCA (python):
      • using dimensionality of extracted feature vector to cluster images
      • should be ported to C++ for the final project
    • SSIM (python):
      • uses similarity measure wihout relying on feature vector to cluster images
      • should be ported to C++ for the final project
    • image-generation (python):
      • will be used to generate images for the clustering
      • performs: pdb aggregation, pdb generation, image generation
    • comvi:
      • C++ project that will be used as a 2D rendered as a plugin in Megamol

• Images:

  • Contains Images to sample
  • Delete when image-generation is up and functional for arbitraty pdb datasets

Already Implemented:

• feature extraction

• similarity measure (SSIM, cosine (with feature vector))

• clustering with DBSCAN

• dim. reduction

• image generation

• image exploration in clusterview