Depict Shape Forrester Cole Kevin Sanik Doug DeCarlo Adam Finkelstein Thomas Funkhouser Szymon Rusinkiewicz Manish Singh Rutgers Princeton Line drawings Matisse 1932 Flaxman 1805 US Patent 378973 ID: 157759
Download Presentation The PPT/PDF document "How Well Do Line Drawings" is the property of its rightful owner. Permission is granted to download and print the materials on this web site for personal, non-commercial use only, and to display it on your personal computer provided you do not modify the materials and that you retain all copyright notices contained in the materials. By downloading content from our website, you accept the terms of this agreement.
Slide1
How Well Do Line Drawings Depict Shape?
Forrester ColeKevin SanikDoug DeCarloAdam FinkelsteinThomas FunkhouserSzymon RusinkiewiczManish Singh
Rutgers
PrincetonSlide2
Line drawings
[Matisse 1932]
[Flaxman 1805]
[US Patent 378,973]Slide3
Line drawings
Occluding Contours
Sharp creasesSlide4
Line drawings
Ridges and Valleys
Suggestive Contours
[DeCarlo et al 2003]
Apparent Ridges
[Judd et al 2007]
Occluding Contours
Sharp creasesSlide5
Assessing Line Drawings
GoalsArtistry, abstraction, etc.Leads to accurate perception of shapeSlide6
Assessing Line Drawings
GoalsArtistry, abstraction, etc.Leads to accurate perception of shapeMethodologyQualitative (examples, comparison to artists)Quantitative comparison to artists’ drawingsDirect measurement of perceived shapeSlide7
Comparing models
Ridges and Valleys
Suggestive Contours
Apparent RidgesSlide8
Comparing models
Ridges and Valleys
Suggestive Contours
Apparent RidgesSlide9
Comparing models
Ridges and Valleys
Suggestive Contours
Apparent RidgesSlide10
Comparing models to artists
©
Estate of Roy Lichtenstein
Suggestive contours and suggestive highlights [DeCarlo and Rusinkiewicz 2007]
“Golf Ball” [Lichtenstein 1962]Slide11
Comparing models to artists
Argument for ridge-like features
[Judd et al. 2007]
[Brancusi 1910]
[Matisse 1932]Slide12
Comparing models to artists
Comparisons to drawings made under controlled conditions [Cole et al. 2008]
artist drawing
apparent ridges
suggestive contours
…Slide13
Comparing models to artists
Comparisons to drawings made under controlled conditions [Cole et al. 2008]
artist drawing
apparent ridges
suggestive contours
…
d
line(
, )
renderingartist drawingSlide14
rendering
Comparing shapes
3D
d
3D
(
,
)
perceived shape
original shapeSlide15
Measuring perceived shape
Local measurements of shape geometry
Gauge figure adjustment
[
Koenderink et al. 1992]Slide16
Measuring perceived shape
Local measurements of shape geometryGauge figure adjustment[Koenderink et al. 1992]Studied shaded surfaces
and one artist line drawing[Koenderink et al. 1996] Slide17
Questions
Do artist and CG drawings effectively convey shape?how accurate are they?how do they compare to a shaded rendering?Do different viewers perceive the same shape?When are particular line types most effective?Slide18
Study Methodology
Measure perceptsBoth artist and CG drawingsRange of modelsMany participantsCompare against ground truth3D shape and shaded imageAccuracy and precision Slide19
Orienting a GaugeSlide20
Example SessionSlide21
Study Setup
All 12 models from [Cole et al. 2008]Slide22
Shaded
R. and V.
Sug
. C.
App. R.
Artist’s
Study Setup
6 styles
x
12 models
- 2 duplicates
= 70 prompts
ContoursSlide23
Shaded
R. and V.
Sug
. C.
App. R.
Artist’s
Study Setup
6 styles
x
12 models
- 2 duplicates
= 70 prompts
ContoursSlide24
Study Setup
70
x 90
gauges / prompt
x 2
s
ettings / opinion
prompts
≈ 100,000
settings
x 8
opinions / gaugeSlide25
Study Setup
x 4seconds / setting
111
hours
70
x 90
gauges / prompt
x 2
settings / opinion
prompts
≈ 100,000
settings
x 8
opinions / gaugeSlide26
So Much Data…
Amazon Mechanical Turk to the rescue!Turker sets 60 gauges, gets paid $0.20Efficient even after throwing away garbage“Garbage” is inconsistent dataAbout 80% of data is consistentSlide27
Data Collection275,000 gauge settings
4 models x 180 gauges + 8 models x 90 gaugesEach gauge 9 to 29 opinions, average 15560 different people
AssignmentsCompleted
# ParticipantsSlide28
Global Accuracy
Error from ground (accuracy)Slide29
Global Accuracy
Error from ground (accuracy)
Distribution of errors for shadedSlide30
Finding:
On average, turkers did a good jobSlide31
Aggregating Per-Gauge Data
What is the most representative direction?“Mean” is most obvious choice“Median” more robust to outliers
mean
medianSlide32
Global Accuracy and Precision
Error from Ground (Accuracy)
Error from Median (Precision)Slide33
Results:Precision greater than accuracy
Accuracy varies with style, precision does notSlide34
Finding:
Peoples’ interpretations of shape are similar, even when those interpretations do not match ground truth.Slide35
Question:
Where are the errors?Slide36
Accuracy by Model
Avg. Error
(degrees)Slide37
Accuracy by Model
Avg. Error (degrees)Slide38
Gauge Visualization: Screwdriver
Contours Only
Artist’s Drawing
0
90
Error (deg.)
180 gaugesSlide39
Local Errors: Screwdriver
Contours Only
Artist’s Drawing
15 gauges, 5 pixel spacing
0
90
Error (deg.)Slide40
Curvature
: Screwdriver
Contours Only
Artist’s Drawing
Contours Only
Artist’s Drawing
Ground Truth
Zero CurvatureSlide41
Gauge Visualization: Flange
Suggestive Contours
180 gauges
Ridges and Valleys
0
90
Error (deg.)Slide42
Local Errors: Flange
Suggestive Contours
Ridges and Valleys
15 gauges, 5 pixel spacing
0
90
Error (deg.)Slide43
Curvature: Flange
Suggestive Contours
Ridges and Valleys
Ground Truth
R. and V.
Sug
. ContoursSlide44
Gauge Visualization:
Rockerarm
Apparent Ridges
90 gauges
Ridges and Valleys
0
90
Error (deg.)Slide45
Non-Local Effects: Rockerarm
-90
90Error Difference (deg)
Worse than RV
Better than RV
Apparent RidgesSlide46
Conclusions
Different people interpret drawings similarlySome drawings almost match shaded imagesLine drawings vary in effectivenessErrors can be traced to specific linesSlide47
Future Work
More analysis of collected dataTowards interpretation model for linesFurther investigation of study methodologyData available at:http://lineshape.cs.princeton.eduSlide48
Thank You
Thanks to Andrew Van Sant and John WilderSupport by NSF grants CCF-0347427, CCF-0541185, CCF- 0702672, CCF-0702580, IIS-0511965, and IIS-0612231, and GoogleModels from Aim@Shape, VAKHUN, and Cyberware
Data available at:http://lineshape.cs.princeton.eduSlide49
Global Accuracy and Precision
Before bas-relief fitting
Error from Ground (Accuracy)
Error from Median (Precision)Slide50
Global Accuracy and Precision
After bas-relief fitting
Error from Ground (Accuracy)
Error from Median (Precision)Slide51
Bas-Relief Ambiguity
Ambiguity in perception of shaded shapes [Koenderink 2001]
=
?Slide52
Line Drawing Ambiguity
Line drawings are even less constrained
=
?Slide53
Gauge Visualization: Flange, #2
Artist’s Drawing
180 gauges
Ridges and Valleys
0
90
Error (deg.)Slide54
Extra Line: Flange
-90
90
Error Difference (deg)
Worse than RV
Better than RV
Artist’s DrawingSlide55
Non-Local Effects: Flange
-90
90
Error Difference (deg)
Worse than RV
Better than RV
Apparent Ridges