Presentation by: Maria Veiga Zrich Introduction Motivation What is out there In depth projects Painting fool eDavid Conclusion Zrich Artists perspective & questions: Generative art introduces complexity to an art piece; Often used as a tool, thus: What techniques can be used? How to introduce randomness to art?
Scientific perspective & questions: Interest in setting up an autonomous creative agent How to formalize a creative process? How to create an autonomous creative agent? How to teach a robot to paint? Zrich Timeline Mozart 18th century Alfred Jensen: The Apex is Nothing (1960) Ellsworth Kelly: Four Black and Whites (1957) John Cage: Music of Changes (1951) The Painting Fool (2006) eDavid (2009) BNJMN (2013) AARON (1973) Zanelle (2012) Vangobot (2012) Frieder Nake (1963) Zrich Timeline Mozart 18th century Alfred Jensen: The Apex is Nothing (1960) Ellsworth Kelly: Four Black and Whites (1957) John Cage: Music of Changes (1951) The Painting Fool (2006) eDavid (2009) BNJMN (2013) AARON (1973) Zanelle (2012) Vangobot (2012) Frieder Nake (1963) Zrich Started in the Computational Creativity Research group at Imperial College
In 2006, the software was given the name Painting Fool
Objective is to build a software system that is one day taken seriously as a creative artist Zrich
The question is: how to build software that exhibits behaviours deemed creative in people? Zrich As an AI project Simulation of paint strokes on a canvas Aware of emotions of the subject Cultural context Critical appraisal of its own work Zrich As an AI project Simulation of paint strokes on a canvas Aware of emotions of the subject Cultural context Critical appraisal of its own work Non-Photorealistic Rendering (NPR): By producing software that simulates natural media Allow artists to teach the software: Allow artists, designers, and anyone else to train the sofware in aspects of its processing (from analysing digital photographs to constructing and painting scenes)
Zrich As an AI project Simulation of paint strokes on a canvas Aware of emotions of the subject Cultural context Critical appraisal of its own work Mapping of painting style to emotions: Mapping composed of researchs group choices and also randomly generated Detect emotion in images and videos: Emotion detection software that identifies: 1) The frame that shows more emotion 2) Locations of the facial features 3) The emotion expressed by the subject (Using methods by Valstar and Pantic)
Zrich As an AI project Simulation of paint strokes on a canvas Aware of emotions of the subject Cultural context Critical appraisal of its own work Show intention in the paintings: Scan the internet for text and image resources for production of artwork. 1) Software downloads news headlines 2) Use a text rank algorithm to find most important nouns 3) Use the keywords to scan image repositories (Google, Flickr) 4) Create a piece by collage of selected pictures and using non-photorealistic rendering software
Zrich As an AI project Simulation of paint strokes on a canvas Aware of emotions of the subject Cultural context Critical appraisal of its own work Still an active area of research: How to formalise the process of appraisal of artwork? Implemented abilities for the software to calculate objective measures of abstract art, e.g: location of symmetries BUT hard to train a software to appreciate works of art without being an extension of its creators subjective views Zrich Emotionally aware painter Demonstrating intent Travelling salesman dancers Zrich Zrich Objective: Build a robot able to paint with real color on a real canvas.
How? Commercial robotic arm + Visual System + Smart Algorithms Zrich Objective: Build a robot able to paint with real color on a real canvas.
How? Commercial robotic arm + Visual System + Smart Algorithms -one-arm industry robot used in car manufacturing modified for painting
- several adapters for different mediums Zrich Objective: Build a robot able to paint with real color on a real canvas.
How? Commercial robotic arm + Visual System + Smart Algorithms Zrich Objective: Build a robot able to paint with real color on a real canvas.
How? Commercial robotic arm + Visual System + Smart Algorithms Painting as an optimization process consider spatial division, color and content composition. Iterative operation scheme to produce color distributions.
2 ways to compute the brush stroke: - Pre-define a set of stroke candidates - Dynamically generated strokes Compute image gradient using a Sobel operator. Randomly generate initial positions. For each position, a path using Line Integral Convolution of the gradient vector field is generated and then Runge Kutta numerical integration is performed.
Zrich Objective: Build a robot able to paint with real color on a real canvas.
How? Commercial robotic arm + Visual System + Smart Algorithms
Semantic Hints: - Use different techniques for different parts of a picture - Automatically segment the picture - Choose which method dynamic or pre-defined brush strokes best minimizes the distance between canvas image C and target image T.
Zrich Zrich Pre-defined brush stroke vs dynamic brush strokes Zrich Traditionally, artists use robotics & generative methods as way to enhance their work Whereas scientists seem to also be looking into a formalisation of the creative process (Computational Creative Theory) However, collaborations between artists and scientists are increasing boundaries arent so clear anymore (Vangobot, Zanelle, Computational Aesthetical Evaluation)
The intention is not to replace artists but software can perform tasks that people cannot perform e.g: the calculation of where to place thousands of strokes at thousands of time steps
Zrich Rise of highly costumizable designs and art pieces Exploration of new ways of aesthetics
Other projects that were influenced by this research field
Generative code Procedural gaming Formalised criteria to judge Goodness of an art piece