On Realism of Architectural Procedural Models Jan Bene š, Tom Kelly, Filip Děchtěrenko , Jaroslav Křivánek, Pascal Müller
Quick Quiz 2 Jan Bene š - On Realism of Architectural Procedural Models
Quick Quiz Computer Generated Photograph 3 Jan Bene š - On Realism of Architectural Procedural Models
Overview Experiment Procedural Modeling Answers & & Data Acquisition Realism Questions Previous Work 4 Jan Bene š - On Realism of Architectural Procedural Models
Procedural Modeling • Procedural modeling – Tool for artists [Palubicki09] – Or fully automatic – Wide range of outputs • Wide use [Independence Day] 5 Jan Bene š - On Realism of Architectural Procedural Models
Procedural Modeling • Procedural modeling – Tool for artists [Palubicki09] – Or fully automatic – Wide range of outputs • Wide use [Independence Day] 6 Jan Bene š - On Realism of Architectural Procedural Models
Realism Realism Plausibility Immersion • Model & material & display • Should be studied • For procedural modelling [Velocipedia, Gianluca Gimini] 7 Jan Bene š - On Realism of Architectural Procedural Models
Realism Realism Plausibility Immersion • Model & material & display • Should be studied • For procedural modelling [Velocipedia, Gianluca Gimini] 8 Jan Bene š - On Realism of Architectural Procedural Models
What we did Procedural architecture 9 Jan Bene š - On Realism of Architectural Procedural Models
What we did User Study 10 Jan Bene š - On Realism of Architectural Procedural Models
Question #1 (1) Can people tell procedurally generated buildings from real ones? 11 Jan Bene š - On Realism of Architectural Procedural Models
Question #2 (2) Is realism carried in the detail or in the larger structure? 12 Jan Bene š - On Realism of Architectural Procedural Models
Question #2 (2) Is realism carried in the detail or in the larger structure? 13 Jan Bene š - On Realism of Architectural Procedural Models
Question #2 (2) Is realism carried in the detail or in the larger structure? 14 Jan Bene š - On Realism of Architectural Procedural Models
Question #3 I think… (3) What factors do the users think influence the perception of realism? 15 Jan Bene š - On Realism of Architectural Procedural Models
Questions (1) Can people tell procedurally generated buildings from real ones? (2) Is realism carried in the detail or in the larger structure? (3) What factors do the users I think… think influence the perception of realism? 16 Jan Bene š - On Realism of Architectural Procedural Models
Previous Work 17
Previous Work - Realism “There is no dirt, no dust, no fingerprints on the glass table…simply too beautiful, too clean and polished…” [Reinhard13] 18 Jan Bene š - On Realism of Architectural Procedural Models
Previous Work - Realism CG2Real [Johnson11] 19 Jan Bene š - On Realism of Architectural Procedural Models
Previous Work - Realism [Rademacher01] 20 Jan Bene š - On Realism of Architectural Procedural Models
Previous Work - Buildings • Building generation – Grammar [Stiny80, Wonka03, Muller06, Schwarz15] – Data [Fan16] – Sketch [Nishida16] – Predefined parts [Kalogerakis12, Talton12] 21 Jan Bene š - On Realism of Architectural Procedural Models
Previous Work - Miscellaneous • Machine Learning – What makes Paris look like Paris [Doersch12] – Architectural Style Recognition [Mathias11] • Image Quality & Similarity – Visible Differences Predictor [Daly92] – Visual Equivalence and Aggregates [Ramanarayanan07,08] – Structural Similarity - SSIM [Wang04] 22 Jan Bene š - On Realism of Architectural Procedural Models
Experimental Setup 23
Datasets Favela Medieval Paris Venice 24 Jan Bene š - On Realism of Architectural Procedural Models
Datasets Computer Generated Photos Venice Favela Medieval Paris 25 Jan Bene š - On Realism of Architectural Procedural Models
Classification Screen X X 26 Jan Bene š - On Realism of Architectural Procedural Models
Blurs 1px – no blur 7px blur 13px blur 55px blur 25px blur 32px blur 27 Jan Bene š - On Realism of Architectural Procedural Models
Summary TRAINING 28 Jan Bene š - On Realism of Architectural Procedural Models
Summary TRAINING Venice-55px Medieval-55px Paris-55px Favela-55px 29 Jan Bene š - On Realism of Architectural Procedural Models
Summary TRAINING Venice-55px Venice-37px Medieval-55px Medieval-37px Paris-55px Paris-37px Favela-55px Favela-37px 30 Jan Bene š - On Realism of Architectural Procedural Models
Summary TRAINING Venice-1px Venice-55px Venice-37px … Medieval-55px Medieval-1px Medieval-37px Paris-1px Paris-55px Paris-37px Favela-55px Favela-37px Favela-1px 31 Jan Bene š - On Realism of Architectural Procedural Models
Summary TRAINING Venice-1px Venice-55px Venice-37px … Medieval-55px Medieval-1px Medieval-37px Paris-1px Paris-55px Paris-37px Favela-55px Favela-37px Favela-1px DEBRIEFING 32 Jan Bene š - On Realism of Architectural Procedural Models
Data preparation 33
Photographs 34 Jan Bene š - On Realism of Architectural Procedural Models
Renders 35 Jan Bene š - On Realism of Architectural Procedural Models
Data Characteristics 300px 300px 36 Jan Bene š - On Realism of Architectural Procedural Models
Data Characteristics 300px Choose subset 300px 37 Jan Bene š - On Realism of Architectural Procedural Models
Quick Peek at Results 38
Quick Results random choice = 0.5 39 Jan Bene š - On Realism of Architectural Procedural Models
Quick Results random choice = 0.5 40 Jan Bene š - On Realism of Architectural Procedural Models
Quick Results Medieval Worst Dataset random choice Favela Best Dataset 41 Jan Bene š - On Realism of Architectural Procedural Models
Q1: CG vs Real 42
Q1: CG vs Real • H DIFF : participants can tell photographs and generated buildings apart. – Random choice = 0.5 accuracy – No blur, overall – p <.001 – Can accept H DIFF – Also true for each dataset • Also true @ 55px 43 Jan Bene š - On Realism of Architectural Procedural Models
Q1: CG vs Real • H DIFF : participants can tell photographs and generated buildings apart. – Random choice = 0.5 accuracy – No blur, overall – p <.001 – Can accept H DIFF – Also true for each dataset • Also true @ 55px 44 Jan Bene š - On Realism of Architectural Procedural Models
Q2: Details vs Structure 45
Q2: Details vs Structure • H SCALE : the detail that allows participants to tell photographs and generated images apart is present at various scales. • ANOVA 46 Jan Bene š - On Realism of Architectural Procedural Models
What’s ANOVA? Independent variable Independent Dependent Black box variable variable Independent variable 47 Jan Bene š - On Realism of Architectural Procedural Models
What’s ANOVA? Blur Size Original Black box Accuracy image class Dataset 48 Jan Bene š - On Realism of Architectural Procedural Models
What’s ANOVA? • ANOVA – Linear model – Effects of independent variables 49 Jan Bene š - On Realism of Architectural Procedural Models
Significant Effects • Significant effects – Blur, p <.001 – Dataset, p <.001 • class, p =.11 • Blur, p <0.001 =>H SCALE accepted 50 Jan Bene š - On Realism of Architectural Procedural Models
H SCALE • Significant effects – Blur, p <.001 – Dataset, p <.001 • Blur, p <.001 =>H SCALE accepted 51 Jan Bene š - On Realism of Architectural Procedural Models
More on Blur • Significant effects ? – Blur, p <.001 – Dataset, p <.001 • Blur, p <.001 =>H SCALE accepted 52 Jan Bene š - On Realism of Architectural Procedural Models
Post-Hoc - Blur Significant Not significant 53 Jan Bene š - On Realism of Architectural Procedural Models
Blurs 1px – no blur 7px blur 13px blur 25px blur 32px blur 55px blur H SCALE : the detail that allows participants to tell photographs and generated images apart is present at various scales. 54 Jan Bene š - On Realism of Architectural Procedural Models
Two- & Three-Way Interactions Blur Size • Three factors • Interactions • More in paper • Significant • Not Significant Original Image Class Dataset 55 Jan Bene š - On Realism of Architectural Procedural Models
Q2: Summary • Blur – p <.001, significant factor – =>H SCALE accepted – Realism at different scales • Exploratory Analysis – Further insight – Two- & Three-way interactions – Mostly significant 56 Jan Bene š - On Realism of Architectural Procedural Models
Q3: What users thought 57
Participants • 52 total – 11 female – 24.3 years old 58 Jan Bene š - On Realism of Architectural Procedural Models
Recommend
More recommend