Phase 1: Grounding Research and Concept Ideation acla5096 A lot of - - PowerPoint PPT Presentation

“A lot of people think that caricature is about picking out someone’s worst feature and exaggerating it as far as you can. That’s wrong. Caricature is basically finding the truth. And then you push the truth.” - Jason Seiler

Phase 1: Grounding Research and Concept Ideation

acla5096

The Caricature

Derived from the Italian word caricare - to charge or load Meaning: a loaded portrait Caricature can be symbolic or pictorial hyperbole. They can be insulting or complimentary. Most commonly of politicians or celebrities.

The plumb-pudding in danger' (William Pitt; Napoleon Bonaparte) by James Gillray, 1805 Ancient Pompeiian graffiti caricature of a politician

The Caricature

What makes them a success is the artist’s ability to zero in on the most distinctive aspect of somebody Filter out details to get a few defining features that are exaggerated Works better drawing off memory rather than a live model, at risk of producing a too realistic image Strong example of a creative act that is subject to preconceptions of beauty, political views, comedic purpose, motivations, stereotypes etc - inherently human characteristics

Facial Recognition

Caricature is similar to the way humans encode faces and remember them - record the most defining features that allow faces to be distinguished easily Ramachandran[1] suggests the peak shift effect in which the brain favours an exaggerated concept more than a non-exaggerated version. Supported by Leopold et al.’s findings that the face-recognising neurons seem to strongly respond to the caricature version of the same faces.[1]

[1] Informatics Engineering and Information Science, Part II: International Conference, ICIEIS, Part 2 of Informatics Engineering and Information Science: International Conference : Proceedings, Azizah Abd Manaf, Exploring Particle-Based Caricature Generations - Somnuk Phon-Amnuaisuk, 2011

Computer Generated Caricature

2 techniques: line drawing or image warping Process automated by variance exaggeration which exaggerates facial components that deviate from the norm, or example based generation that exaggerate components according to provided templates

Computer Generated Caricature

Susan E. Brennan was the pioneer of the Caricature Generator[2]. Altered the line drawing of a face by exaggerating metric differences between the subject and prototype norm.

[2] Caricature Generator: The Dynamic Exaggeration of Faces by Computer, Susan E. Brennan, Leonardo, Vol. 18, No. 3 (1985), pp. 170-178, Published by: The MIT Press

Computer Generated Caricature

Akleman[3] employed a deformation based approach using morphing/warping.

[3] E Aklemen, Making Caricatures with Morphing, Proceedings of the International Conference on Computer graphics and Interactive Technique ACM SIGGRAPH 1997, p 145

Computer Generated Caricature

Somnuk Phon-Amnuaisuk[1] explores an interactive particle based technique that relies of user feedback to explore the caricature face space. Facial feature points are represented as particle and their movements are used to incrementally warps a given face until the desired exaggerations are achieved.

[1] Informatics Engineering and Information Science, Part II: International Conference, ICIEIS, Part 2 of Informatics Engineering and Information Science: International Conference : Proceedings, Azizah Abd Manaf, Exploring Particle-Based Caricature Generations - Somnuk Phon-Amnuaisuk, 2011

Idea Generation

• What would happen if a robot was asked to create a caricature?
• How would you create a caricature free from human preconceptions,

prejudices, stereotypes, political views, beauty ideals? Could these be intrinsically human things be replicated in code?

• Exploration of how we see ourselves/each other by removing societal

preconceptions

• Exploration of how a robot would identify distinguishing features for facial

recognition

• Applications in facial recognition? Aid in learning faces? Identifying facial

expressions and their corresponding emotions?

Concept 1 - Comparative Caricature

An exploration of how robots see us and distinguish us from one another.

• Interaction Scenario:
• 1. Two people enter view of camera and have portrait taken
• 2. Robot identifies facial features, creates norm “average” or morph of 2 faces
• 3. Analyses differences between each face and selects features to emphasise
• 4. Robot draws portrait of each user with distorted features
• 5. User repeats process with another partner and compare the caricatures

generated

Concept 1 - Comparative Caricature

• 1. Two people enter view of camera and have portrait taken
• 2. 2D image is created for comparison of 2 faces
• 3. Facial features are mapped as points. Distances and

sizes of features are calculated and most significant differences are selected for exaggeration.

• 4. Caricature is drawn by distorting facial features

(placement of eyes, size of nose, width of mouth etc) Thick felt tipped pen used to achieve stylised effect.

Concept 2 - Deviation from Norm

An exploration of how a human and robot can collaborate to create a caricature.

• Interaction Scenario
• 1. ‘Norm’ face is calculated from average of large sample of portrait images
• 2. Image of user’s face is captured by camera
• 3. Image is analysed to calculate metrics of facial features, and how much they

deviate from the ‘norm’

• 4. User is presented with several variations of distortion (different facial features’

deviations chosen to be exaggerated e.g. size, placement, symmetry)

• 5. User selects favourite variation to proceed
• 6. Robot continues to exaggerate or distort this selected feature and present

user with iterative variations

• 7. Selection process is repeated until user is happy with resulting distortion
• 8. Robot draws this image as a stylised caricature

Concept 2 - Deviation from Norm

• 1. User sits down in a typical live model caricature setting.

Camera takes image of user’s face.

• 2. Image is analysed to calculate metrics of facial features,

and how much they deviate from the ‘norm’

• 3. User is presented with a set of on-screen variations each

exaggerating a different facial feature. User selects their favourite distortion to proceed.

• 4. After several iterations user can select that they are

happy with a final variation and the robot draws this as a stylised caricature