Pattern-Driven Navigation in 2D Multiscale Visualizations with Scalable Insets

Scalable Insets is a novel technique for pattern-guided navigation in multiscale visualizations such as geographics maps, gigapixel images, or large matrices like genome interaction maps.

These slides are from my paper talk at IEEE VIS InfoVis 2019.

Paper: https://vcg.seas.harvard.edu/pubs/scalable-insets
Video: https://youtu.be/7Bn4mNLl3WQ
Code etc: https://scalable-insets.lekschas.de

Transcript

1. Fritz Lekschas @flekschas

   Harvard University, School of Engineering
   Michael Behrisch, Benjamin Bach, Peter Kerpedjiev,

   Nils Gehlenborg, and Hanspeter Pfister
   IEEE VIS

   Oct 23rd 2019
   PATTERN-DRIVEN NAVIGATION

   IN 2D MULTISCALE VISUALIZATIONS

   WITH SCALABLE INSETS
   

   View Slide

2. Fritz Lekschas @flekschas

   Harvard University, School of Engineering
   Michael Behrisch, Benjamin Bach, Peter Kerpedjiev,

   Nils Gehlenborg, and Hanspeter Pfister
   IEEE VIS

   Oct 23rd 2019
   PATTERN-DRIVEN NAVIGATION

   IN 2D MULTISCALE VISUALIZATIONS

   WITH SCALABLE INSETS
   SCALABLE INSETS
   

   View Slide

3. View Slide

4. View Slide

5. View Slide

6. View Slide

7. View Slide

8. View Slide

9. View Slide

10. GOALS
    Identify promising regions

    for detailed exploration
    Find pattern instances

    within their context
    Compare patterns

    across scales
    

    View Slide

11. EFFECTIVENESS
    See many insets

    Close to origin
    Minimize occlusion
    

    View Slide

12. EFFECTIVENESS
    See many insets

    Close to origin
    Minimize occlusion
    

    View Slide

13. EFFECTIVENESS
    See many insets

    Close to origin
    Minimize occlusion
    

    View Slide

14. EFFECTIVENESS
    See many insets

    Close to origin
    Minimize occlusion
    See many insets
    

    View Slide

15. EFFECTIVENESS
    See many insets

    Close to origin
    Minimize occlusion
    Close to origin
    

    View Slide

16. EFFECTIVENESS
    See many insets

    Close to origin
    Minimize occlusion
    Minimize occlusion
    

    View Slide

17. SCALABILITY
    Control inset size

    Group insets
    

    View Slide

18. View Slide

19. GIGAPIXEL IMAGES
    Image tiles from The Rio—Hong Kong Connection
    

    View Slide

20. Zoom-invariant size: control
    details and context
    Dynamic leader line:

    avoid clutter

    preserve context
    INSET DESIGN
    

    View Slide

21. Zoom-invariant size: control
    details and context
    Dynamic leader line:

    avoid clutter

    preserve context
    INSET DESIGN
    

    View Slide

22. Simulated annealing
    Minimize:

    inset distance

    inset movement

    inset overlap

    leader-line crossing
    Evaluate on zoom
    INSET PLACEMENT
    

    View Slide

23. Clustering:

    projected distance and area
    in pixel space
    Stability:

    merge and split points differ

    avoid abrupt changes
    INSET GROUPING
    

    View Slide

24. Image data:

    gallery of representatives

    highlight diversity
    Matrices:

    pile of insets

    summary statistics
    INSET AGGREGATION
    

    View Slide

25. 1. User study: Performance comparison in

    frequency estimation, pattern search, and pattern comparison
    on gigapixel images
    Participants: 18
    Techniques:
    Results: · faster pattern search

    · more accurate pattern comparison

    · little overhead for frequency estimation
    EVALUATION
    

    View Slide

26. 6 domain experts: free exploration + think aloud
    Easy-to-learn and useful: picked up in ~1min
    Inner placement:

    guided navigation
    Outer placement:

    pattern search

    & comparisons
    2. USER STUDY
    

    View Slide

27. Explore sparsely-distributed local patterns in context
    Faster pattern search
    Improved navigation
    CONCLUSION
    

    View Slide

28. Thank You!
    PAPER
    vcg.seas.harvard.edu/pubs/scalable-insets
    CODE, SLIDES, ETC.
    scalable-insets.lekschas.de
    CONTACT
    @flekschas
    FUNDING
    CO-AUTHORS
    Michael Behrisch, Benjamin Bach, Peter Kerpedjiev
    Nils Gehlenborg, and Hanspeter Pfister
    is hiring!
    

    View Slide