On the Readability of Boundary Labeling Lukas Barth, Andreas Gemsa, - - PowerPoint PPT Presentation

on the readability of boundary labeling
SMART_READER_LITE
LIVE PREVIEW

On the Readability of Boundary Labeling Lukas Barth, Andreas Gemsa, - - PowerPoint PPT Presentation

On the Readability of Boundary Labeling Lukas Barth, Andreas Gemsa, Benjamin Niedermann, Martin N ollenburg 1 On the Readability of Boundary Labeling L. Barth, A. Gemsa, B. Niedermann, N. N ollenburg Motivation


slide-1
SLIDE 1

On the Readability of Boundary Labeling

  • L. Barth, A. Gemsa, B. Niedermann, N. N¨
  • llenburg

1

Lukas Barth, Andreas Gemsa, Benjamin Niedermann, Martin N¨

  • llenburg

On the Readability of Boundary Labeling

slide-2
SLIDE 2

On the Readability of Boundary Labeling

  • L. Barth, A. Gemsa, B. Niedermann, N. N¨
  • llenburg

2

Motivation

https://de.wikipedia.org/wiki/Auge#/media/File:Eye scheme.svg

slide-3
SLIDE 3

On the Readability of Boundary Labeling

  • L. Barth, A. Gemsa, B. Niedermann, N. N¨
  • llenburg

2

Motivation

Vitreous humour Retina Sclera Optic nerve Lens Iris Pupil Choroid Cornea Zonular fibres Anterior chamber Posterior chamber Ciliary muscle

https://de.wikipedia.org/wiki/Auge#/media/File:Eye scheme.svg

slide-4
SLIDE 4

On the Readability of Boundary Labeling

  • L. Barth, A. Gemsa, B. Niedermann, N. N¨
  • llenburg

2

Motivation

Vitreous humour Retina Sclera Optic nerve Lens Iris Pupil Choroid Cornea Zonular fibres Anterior chamber Posterior chamber Ciliary muscle

Outline: Discussion on mathematical models. First user study on readability of boundary labeling.

https://de.wikipedia.org/wiki/Auge#/media/File:Eye scheme.svg

slide-5
SLIDE 5

On the Readability of Boundary Labeling

  • L. Barth, A. Gemsa, B. Niedermann, N. N¨
  • llenburg

3

Mathematical Models

slide-6
SLIDE 6

On the Readability of Boundary Labeling

  • L. Barth, A. Gemsa, B. Niedermann, N. N¨
  • llenburg

4

Typical Formalization

Given: Rectangle R Points P = {p1, . . . , pn} in R. For each pi ∈ P: label ℓi given by bounding box. p1 p3 p2 p4 R Mouse Elephant Cow Snake ℓ2 = ℓ1 = ℓ3 = ℓ4 =

slide-7
SLIDE 7

On the Readability of Boundary Labeling

  • L. Barth, A. Gemsa, B. Niedermann, N. N¨
  • llenburg

4

Typical Formalization

Given: Rectangle R Points P = {p1, . . . , pn} in R. For each pi ∈ P: label ℓi given by bounding box. p1 p3 p2 p4 R Find: Placement of labels such that labels lie outside of R labels touch the border of R no two labels overlap ∃ simple curve λi in R connecting pi with ℓi. Mouse Elephant Cow Snake ℓ2 = ℓ1 = ℓ3 = ℓ4 = p1 p3 R Elephant Snake Cow Mouse p4 p2 Typicially: crossing free total length minimization

slide-8
SLIDE 8

On the Readability of Boundary Labeling

  • L. Barth, A. Gemsa, B. Niedermann, N. N¨
  • llenburg

4

Typical Formalization

Given: Rectangle R Points P = {p1, . . . , pn} in R. For each pi ∈ P: label ℓi given by bounding box. p1 p3 p2 p4 R Find: Placement of labels such that labels lie outside of R labels touch the border of R no two labels overlap ∃ simple curve λi in R connecting pi with ℓi. Mouse Elephant Cow Snake ℓ2 = ℓ1 = ℓ3 = ℓ4 = p1 p3 R Elephant Snake Cow Mouse p4 p2 Typicially: crossing free total length minimization

Leader

slide-9
SLIDE 9

On the Readability of Boundary Labeling

  • L. Barth, A. Gemsa, B. Niedermann, N. N¨
  • llenburg

5

Typical Formalization

Given: Rectangle R Points P = {p1, . . . , pn} in R. For each pi ∈ P: label ℓi given by bounding box. p1 p3 p2 p4 R Find: Placement of labels such that labels lie outside of R labels touch the border of R no two labels overlap ∃ simple curve λi in R connecting pi with ℓi. Mouse Elephant Cow Snake ℓ2 = ℓ1 = ℓ3 = ℓ4 = p1 p3 R Elephant Snake Cow Mouse p4 p2 Typicially:

Many possible parameters: Label size. Allowed positions of labels. . . . Important parameter: Type of leaders

crossing free total length minimization

slide-10
SLIDE 10

On the Readability of Boundary Labeling

  • L. Barth, A. Gemsa, B. Niedermann, N. N¨
  • llenburg

6

Typical Examples of Leaders

straight line S-shaped L-shaped diagonal

slide-11
SLIDE 11

On the Readability of Boundary Labeling

  • L. Barth, A. Gemsa, B. Niedermann, N. N¨
  • llenburg

6

Typical Examples of Leaders

straight line S-shaped L-shaped diagonal

slide-12
SLIDE 12

On the Readability of Boundary Labeling

  • L. Barth, A. Gemsa, B. Niedermann, N. N¨
  • llenburg

6

Typical Examples of Leaders

straight line S-shaped L-shaped diagonal

slide-13
SLIDE 13

On the Readability of Boundary Labeling

  • L. Barth, A. Gemsa, B. Niedermann, N. N¨
  • llenburg

6

Typical Examples of Leaders

straight line S-shaped L-shaped diagonal

slide-14
SLIDE 14

On the Readability of Boundary Labeling

  • L. Barth, A. Gemsa, B. Niedermann, N. N¨
  • llenburg

6

Typical Examples of Leaders

straight line S-shaped L-shaped diagonal

slide-15
SLIDE 15

On the Readability of Boundary Labeling

  • L. Barth, A. Gemsa, B. Niedermann, N. N¨
  • llenburg

7

Algorithmic Results

Year Paper Leader Type s po do

  • po
  • ther

2004 Bekos et al. × × × 2005 Ali et al. × × 2006 Bekos et al. × 2008 Lin et al. ⋆ ⋆ 2009 Benkert et al. × × Lin et al. × × 2010 Bekos et al. × × Bekos et al. × Lin ⋆ L¨

  • ffler and N¨
  • llenburg

× N¨

  • llenburg et al.

× Year Paper s 2011 Bekos et al. Gemsa et al. × 2012 Fink et al. × 2013 Bekos et al. Kindermann et al. 2014 Huang et al. Kindermann et al. × 2015 L¨

  • ffler et al.

On the Readability of Boundary Labeling

  • L. Barth, A. Gemsa, B. Niedermann, N. N¨
  • llenburg
slide-16
SLIDE 16

On the Readability of Boundary Labeling

  • L. Barth, A. Gemsa, B. Niedermann, N. N¨
  • llenburg

8

Algorithmic Results

Year Paper Leader Type s po do

  • po
  • ther

2004 Bekos et al. × × × 2005 Ali et al. × × 2006 Bekos et al. × 2008 Lin et al. ⋆ ⋆ 2009 Benkert et al. × × Lin et al. × × 2010 Bekos et al. × × Bekos et al. × Lin ⋆ L¨

  • ffler and N¨
  • llenburg

× N¨

  • llenburg et al.

× Year Paper s 2011 Bekos et al. Gemsa et al. × 2012 Fink et al. × 2013 Bekos et al. Kindermann et al. 2014 Huang et al. Kindermann et al. × 2015 L¨

  • ffler et al.

straight line S-shaped L-shaped diagonal

3 9 9 5

Papers on

  • ther types

5

On the Readability of Boundary Labeling

  • L. Barth, A. Gemsa, B. Niedermann, N. N¨
  • llenburg
slide-17
SLIDE 17

On the Readability of Boundary Labeling

  • L. Barth, A. Gemsa, B. Niedermann, N. N¨
  • llenburg

9

User Study on Boundary Labeling

slide-18
SLIDE 18

On the Readability of Boundary Labeling

  • L. Barth, A. Gemsa, B. Niedermann, N. N¨
  • llenburg

10

How to Measure Readability?

Task A Dog Mouse Cat Given: Highlighted label. Find: Corresponding point. Task B Dog Mouse Cat Given: Highlighted point. Find: Corresponding label. Measure: Response time of user. Success rate of user.

slide-19
SLIDE 19

On the Readability of Boundary Labeling

  • L. Barth, A. Gemsa, B. Niedermann, N. N¨
  • llenburg

11

Research Questions

Hypothesis 1a:

straight line S-shaped L-shaped diagonal

better⋆ than

⋆with respect to response time and success rate

Hypothesis 1b:

S-shaped L-shaped diagonal

better⋆ than For dense sets of points

slide-20
SLIDE 20

On the Readability of Boundary Labeling

  • L. Barth, A. Gemsa, B. Niedermann, N. N¨
  • llenburg

12

Research Questions

Hypothesis 2:

S-shaped L-shaped diagonal

better⋆ than For sparse sets of points

straight line

⋆with respect to response time and success rate

slide-21
SLIDE 21

On the Readability of Boundary Labeling

  • L. Barth, A. Gemsa, B. Niedermann, N. N¨
  • llenburg

13

Stimuli

Density of Points: dense N(µ, σsmall) sparse uniform N(µ, σlarge)

slide-22
SLIDE 22

On the Readability of Boundary Labeling

  • L. Barth, A. Gemsa, B. Niedermann, N. N¨
  • llenburg

13

Stimuli

Density of Points: dense N(µ, σsmall) sparse uniform N(µ, σlarge) Size: small: 15 points large: 30 points

slide-23
SLIDE 23

On the Readability of Boundary Labeling

  • L. Barth, A. Gemsa, B. Niedermann, N. N¨
  • llenburg

13

Stimuli

Type of Leaders:

straight line S-shaped L-shaped diagonal

Density of Points: dense N(µ, σsmall) sparse uniform N(µ, σlarge) Size: small: 15 points large: 30 points

slide-24
SLIDE 24

On the Readability of Boundary Labeling

  • L. Barth, A. Gemsa, B. Niedermann, N. N¨
  • llenburg

13

Stimuli

Type of Leaders:

straight line S-shaped L-shaped diagonal

Density of Points: Tasks: dense N(µ, σsmall) sparse uniform Task A Task B

Dog Mouse Cat Dog Mouse Cat

N(µ, σlarge) Size: small: 15 points large: 30 points

slide-25
SLIDE 25

On the Readability of Boundary Labeling

  • L. Barth, A. Gemsa, B. Niedermann, N. N¨
  • llenburg

13

Stimuli

Type of Leaders:

straight line S-shaped L-shaped diagonal

Density of Points: Degree of Difficulty: 3 Levels for choice of leaders. Tasks: dense N(µ, σsmall) sparse uniform Task A Task B

Dog Mouse Cat Dog Mouse Cat

N(µ, σlarge) Size: small: 15 points large: 30 points

slide-26
SLIDE 26

On the Readability of Boundary Labeling

  • L. Barth, A. Gemsa, B. Niedermann, N. N¨
  • llenburg

14

Stimuli

Type of Leaders:

straight line S-shaped L-shaped diagonal

Density of Points: Degree of Difficulty: 3 Levels for choice of leaders. Tasks: dense N(µ, σsmall) sparse uniform Task A Task B 4 × 2 × 4 × 2 × 3 = 144 stimuli per participant!

Dog Mouse Cat Dog Mouse Cat

N(µ, σlarge) Size: small: 15 points large: 30 points

slide-27
SLIDE 27

On the Readability of Boundary Labeling

  • L. Barth, A. Gemsa, B. Niedermann, N. N¨
  • llenburg

15

Examples of Stimuli

L-shaped diagonal straight line S-shaped

15 points, sparse distribution

slide-28
SLIDE 28

On the Readability of Boundary Labeling

  • L. Barth, A. Gemsa, B. Niedermann, N. N¨
  • llenburg

16

Procedure

  • 1. Tutorial
  • 2. 144 stimuli (randomized order)
  • 3. Questionnaire on preferences

6 female 25 male 20-30 years old Two participants stated that they have fundamental knowledge about labeling. 31 Participants: Digital Questionnaire (identical hardware and conditions)

slide-29
SLIDE 29

On the Readability of Boundary Labeling

  • L. Barth, A. Gemsa, B. Niedermann, N. N¨
  • llenburg

17

Results

Dense 101 100 10−1 Response Time: Normalized Significant Difference⋆

⋆Friedman tests with post-hoc Dunn-Bonferroni pairwise comparisons

slide-30
SLIDE 30

On the Readability of Boundary Labeling

  • L. Barth, A. Gemsa, B. Niedermann, N. N¨
  • llenburg

17

Results

Dense 101 100 10−1 Sparse Response Time: Normalized Significant Difference⋆

⋆Friedman tests with post-hoc Dunn-Bonferroni pairwise comparisons

slide-31
SLIDE 31

On the Readability of Boundary Labeling

  • L. Barth, A. Gemsa, B. Niedermann, N. N¨
  • llenburg

17

Results

Dense 101 100 10−1 Sparse Uniform Response Time: Normalized Significant Difference⋆

⋆Friedman tests with post-hoc Dunn-Bonferroni pairwise comparisons

slide-32
SLIDE 32

On the Readability of Boundary Labeling

  • L. Barth, A. Gemsa, B. Niedermann, N. N¨
  • llenburg

17

Results

Dense 101 100 10−1 Sparse Uniform Response Time: Normalized Significant Difference⋆ Summary: better than better than better than

⋆Friedman tests with post-hoc Dunn-Bonferroni pairwise comparisons

slide-33
SLIDE 33

On the Readability of Boundary Labeling

  • L. Barth, A. Gemsa, B. Niedermann, N. N¨
  • llenburg

18

Results

Accuracy: Success rates Significant Difference⋆ Dense 0.6 0.7 0.8 0.9 1.0

⋆Friedman tests with post-hoc Dunn-Bonferroni pairwise comparisons

slide-34
SLIDE 34

On the Readability of Boundary Labeling

  • L. Barth, A. Gemsa, B. Niedermann, N. N¨
  • llenburg

18

Results

Accuracy: Success rates Significant Difference⋆ Dense 0.6 0.7 0.8 0.9 1.0

⋆Friedman tests with post-hoc Dunn-Bonferroni pairwise comparisons

slide-35
SLIDE 35

On the Readability of Boundary Labeling

  • L. Barth, A. Gemsa, B. Niedermann, N. N¨
  • llenburg

18

Results

Accuracy: Success rates Significant Difference⋆ Dense Sparse 0.6 0.7 0.8 0.9 1.0

⋆Friedman tests with post-hoc Dunn-Bonferroni pairwise comparisons

slide-36
SLIDE 36

On the Readability of Boundary Labeling

  • L. Barth, A. Gemsa, B. Niedermann, N. N¨
  • llenburg

18

Results

Accuracy: Success rates Significant Difference⋆ Summary: are better than Dense Sparse Uniform 0.6 0.7 0.8 0.9 1.0

⋆Friedman tests with post-hoc Dunn-Bonferroni pairwise comparisons

slide-37
SLIDE 37

On the Readability of Boundary Labeling

  • L. Barth, A. Gemsa, B. Niedermann, N. N¨
  • llenburg

19

Summary

better than better than better than Response Time: Accuracy: Success rates better than

slide-38
SLIDE 38

On the Readability of Boundary Labeling

  • L. Barth, A. Gemsa, B. Niedermann, N. N¨
  • llenburg

19

Summary

better than better than better than Response Time: Accuracy: Success rates better than Hypothesis 1a (Dense):

straight line S-shaped L-shaped diagonal

better⋆ than

⋆with respect to response time and success rate

Partly confirmed.

slide-39
SLIDE 39

On the Readability of Boundary Labeling

  • L. Barth, A. Gemsa, B. Niedermann, N. N¨
  • llenburg

19

Summary

better than better than better than Response Time: Accuracy: Success rates better than Hypothesis 1b (Dense):

S-shaped L-shaped diagonal

better⋆ than

⋆with respect to response time and success rate

Confirmed.

slide-40
SLIDE 40

On the Readability of Boundary Labeling

  • L. Barth, A. Gemsa, B. Niedermann, N. N¨
  • llenburg

19

Summary

better than better than better than Response Time: Accuracy: Success rates better than Hypothesis 2 (Sparse):

S-shaped L-shaped diagonal

better⋆ than

straight line

⋆with respect to response time and success rate

Confirmed.

slide-41
SLIDE 41

On the Readability of Boundary Labeling

  • L. Barth, A. Gemsa, B. Niedermann, N. N¨
  • llenburg

19

Summary

better than better than better than Response Time: Accuracy: Success rates better than Readability is important, but aesthetics too! How do you rate the appearance of the leader types?

slide-42
SLIDE 42

On the Readability of Boundary Labeling

  • L. Barth, A. Gemsa, B. Niedermann, N. N¨
  • llenburg

19

Summary

better than better than better than Response Time: Accuracy: Success rates better than Readability is important, but aesthetics too! How do you rate the appearance of the leader types? 1 2 3 4 5 6 fail grade

{

German School Grades:

slide-43
SLIDE 43

On the Readability of Boundary Labeling

  • L. Barth, A. Gemsa, B. Niedermann, N. N¨
  • llenburg

19

Summary

better than better than better than Response Time: Accuracy: Success rates better than Readability is important, but aesthetics too! How do you rate the appearance of the leader types? 1 2 3 4 5 6 fail grade

{

German School Grades: 1.8 2.3 3.3 4.6 Average:

slide-44
SLIDE 44

On the Readability of Boundary Labeling

  • L. Barth, A. Gemsa, B. Niedermann, N. N¨
  • llenburg

20

Conclusion

S-shaped L-shaped diagonal

Readability User’s Preference

straight line diagonal L-shaped straight line S-shaped

1 2 3 4 1 2 3 4

Ranking Ranking

slide-45
SLIDE 45

On the Readability of Boundary Labeling

  • L. Barth, A. Gemsa, B. Niedermann, N. N¨
  • llenburg

20

Conclusion

S-shaped L-shaped diagonal

Readability User’s Preference

straight line diagonal L-shaped straight line S-shaped

1 2 3 4 1 2 3 4

Ranking Ranking

What, if there is a figure in the background?

slide-46
SLIDE 46

On the Readability of Boundary Labeling

  • L. Barth, A. Gemsa, B. Niedermann, N. N¨
  • llenburg

20

Conclusion

S-shaped L-shaped diagonal

Readability User’s Preference

straight line diagonal L-shaped straight line S-shaped

1 2 3 4 1 2 3 4

Ranking Ranking

What about other leader types?

slide-47
SLIDE 47

On the Readability of Boundary Labeling

  • L. Barth, A. Gemsa, B. Niedermann, N. N¨
  • llenburg

20

Conclusion

S-shaped L-shaped diagonal

Readability User’s Preference

straight line diagonal L-shaped straight line S-shaped

1 2 3 4 1 2 3 4

Ranking Ranking

What if other criteria are optimized?

slide-48
SLIDE 48

On the Readability of Boundary Labeling

  • L. Barth, A. Gemsa, B. Niedermann, N. N¨
  • llenburg

20

Conclusion

S-shaped L-shaped diagonal

Readability User’s Preference

straight line diagonal L-shaped straight line S-shaped

1 2 3 4 1 2 3 4

Ranking Ranking

Thank you!