A Machine Learning Approach to Make Static Visualizations Interactive

Theoretical (Analytical):

Practical (Implementation):

Literature Work:


During presentations, it is often not feasible to show interactive visualizations of data. Instead, static images of these visualizations are shown. Using machine learning and image recognition, an interactive copy of the visualization could be created and presented to the user. The user is then able to interact with this copy in order to analyze the data. The project will develop a framework for the inspection of static visualizations. The static visualization (print-out, draft, etc.) should be recognized with image processing (machine learning) and overlaid with a digital copy with which the observer can interact with.

Problem Statement

A major drawback of static visualizations is their incapability to adapt to changes in the data or visualization parameters. Machine learning could overcome this problem by creating a dynamic digital clone of the static visualization.


  • Get familiar with deep learning, image recognition, and detection algorithms.
  • Create a framework that is capable of digitalizing a static visualization and extend it by basic interaction functionalities


    • Basic skills in image processing, deep learning etc
    • Advanced programming skills in Java or C#
    • Useful: Git


      • Scope: Bachelor/Master
      • 6 Month Project
      • 3/6 Month Thesis



          • Kim, T., Saket, B., Endert, A., & MacIntyre, B. (2017). VisAR: Bringing Interactivity to Static Data Visualizations through Augmented Reality. arXiv preprint arXiv:1708.01377.
          • Saenz, M., Baigelenov, A., Hung, Y. H., & Parsons, P. Reexamining the cognitive utility of 3D visualizations using augmented reality holograms. IEEE VIS 2017.