Publications

@article{Berio2025NeuralAbstraction,

title = {Neural Image Abstraction using Long Smoothing B-Splines},

author = {Daniel Berio and Michael Stroh and Sylvain Calinon and Frederic Fol Leymarie and Oliver Deussen and Ariel Shamir},

url = {https://dl.acm.org/doi/abs/10.1145/3763345},

doi = {10.1145/3763345},

year  = {2025},

date = {2025-01-01},

urldate = {2025-01-01},

journal = {ACM Transactions on Graphics},

volume = {44},

number = {6},

pages = {1–11},

abstract = {We integrate smoothing B-splines into a standard differentiable vector graphics (DiffVG) pipeline through linear mapping, and show how this can be used to generate smooth and arbitrarily long paths within image-based deep learning systems. We take advantage of derivative-based smoothing costs for parametric control of fidelity vs. simplicity tradeoffs, while also enabling stylization control in geometric and image spaces. The proposed pipeline is compatible with recent vector graphics generation and vectorization methods. We demonstrate the versatility of our approach with four applications aimed at the generation of stylized vector graphics: stylized space-filling path generation, stroke-based image abstraction, closed-area image abstraction, and stylized text generation.},

keywords = {},

pubstate = {published},

tppubtype = {article}

}

@article{Stroh2025CGF,

title = {Using Saliency for Semantic Image Abstractions in Robotic Painting},

author = {Michael Stroh and Patrick Paetzold and Daniel Berio and Rebecca Kehlbeck and Frederic Fol Leymarie and Oliver Deussen and Noura Faraj},

url = {https://onlinelibrary.wiley.com/doi/full/10.1111/cgf.70259},

doi = {10.1111/cgf.70259},

year  = {2025},

date = {2025-01-01},

journal = {Computer Graphics Forum},

volume = {44},

number = {7},

pages = {e70259},

abstract = {We present an adaptive, semantics-based abstraction approach that balances aesthetic quality and structural coherence within the practical constraints of robotic painting. We apply panoptic segmentation with color-based over-segmentation to partition images into meaningful regions aligned with semantic objects, while providing flexible abstraction levels. Automatic parameter selection for region merging is enabled by semantic saliency maps, derived from Out-of-Distribution segmentation techniques in combination with machine learning methods for feature detection. This preserves the boundaries of salient objects while simplifying less prominent regions. A graph-based community detection step further refines the abstraction by grouping regions according to local connectivity and semantic coherence. The runtime of our method outperforms optimization-based image vectorization methods, enabling the efficient generation of multiple abstraction levels that can serve as hierarchical layers for robotic painting. We demonstrate the quality of our method by showing abstraction results, robotic paintings with the e-David robot, and a comparison to other abstraction methods.},

keywords = {},

pubstate = {published},

tppubtype = {article}

}

@article{nokey,

title = {Robotic art: An anthropomorphic investigation of embodiment},

author = {Francesca Mamino and Rebecca Chamberlain and Daniel Berio and Frederic Fol Leymarie and Sofia Reina},

doi = {10.1163/22134913-20250001},

issn = {2213-4905},

year  = {2024},

date = {2024-11-15},

journal = {Art and Perception},

volume = {13},

issue = {1},

abstract = {Previous research suggests that people prefer abstract line drawings made with human-like kinematics (Chamberlain et al., 2021), implying an embodied aesthetic experience tied to the artist's motion during creation (Freedberg & Gallese, 2007). It has also been found that watching a robotic arm drawing also has a positive effect on the aesthetic evaluation of the resulting drawings, again suggesting some role of embodiment in aesthetic perception (Chamberlain et al., 2018). Taking these findings into account, the current study investigates the influence of embodiment on aesthetic evaluation in drawings made using robotics. By systematically modulating the human-like features of robotic systems we can investigate whether human resemblance plays a role in aesthetic responses to robotic drawings. Specifically, we manipulated two core aspects of human resemblance: 1. Movement (the degree of humanness of the movement), and 2. Anthropomorphism (the degree of human resemblance of the machine to a human arm) in a between-subjects design with four conditions. Participants either: 1. Watched a video of a robotic arm with human-like kinematics (human movement and high anthropomorphism) 2. Watched a video of a robotic arm with uniform kinematics (Machine movement with high anthropomorphism) 3. Watched a video of an Axiplot machine with uniform kinematics (Machine movement with low anthropomorphism) 4. Read a piece of text with some information about the machines (Control). Participants were then asked to rate a series of drawings made by the robotic systems under investigation, alongside some individual difference measures of art preference and attitudes to robotics. Data collection is currently ongoing (target sample size = 160). Our hypothesis is that the conditions including human movement and anthropomorphism will lead to higher aesthetic evaluation of the resulting drawings relative to conditions without either human kinematics or anthropomorphic robotic forms.},

note = {Poster presented at VSAC 2024 in Aberdeen, Scotland, August 22-24.

Abstract published in a special issue of Art & Perception: },

keywords = {},

pubstate = {published},

tppubtype = {article}

}

@article{10.1145/3592123,

title = {Word-As-Image for Semantic Typography},

author = {Shir Iluz and Yael Vinker and Amir Hertz and Daniel Berio and Daniel Cohen-Or and Ariel Shamir},

url = {https://doi.org/10.1145/3592123},

doi = {10.1145/3592123},

issn = {0730-0301},

year  = {2023},

date = {2023-07-01},

journal = {ACM Trans. Graph.},

volume = {42},

number = {4},

publisher = {Association for Computing Machinery},

address = {New York, NY, USA},

abstract = {A word-as-image is a semantic typography technique where a word illustration presents a visualization of the meaning of the word, while also preserving its readability. We present a method to create word-as-image illustrations automatically. This task is highly challenging as it requires semantic understanding of the word and a creative idea of where and how to depict these semantics in a visually pleasing and legible manner. We rely on the remarkable ability of recent large pretrained language-vision models to distill textual concepts visually. We target simple, concise, black-and-white designs that convey the semantics clearly. We deliberately do not change the color or texture of the letters and do not use embellishments. Our method optimizes the outline of each letter to convey the desired concept, guided by a pretrained Stable Diffusion model. We incorporate additional loss terms to ensure the legibility of the text and the preservation of the style of the font. We show high quality and engaging results on numerous examples and compare to alternative techniques. Code and demo will be available at our project page.},

keywords = {},

pubstate = {published},

tppubtype = {article}

}

@article{robotics12030072,

title = {Robotic Writing of Arbitrary Unicode Characters Using Paintbrushes},

author = {David Silvan Zingrebe and Jörg Marvin Gülzow and Oliver Deussen},

url = {https://www.mdpi.com/2218-6581/12/3/72},

doi = {10.3390/robotics12030072},

issn = {2218-6581},

year  = {2023},

date = {2023-01-01},

journal = {Robotics},

volume = {12},

number = {3},

abstract = {Human handwriting is an everyday task performed regularly by most people. In the domain of robotic painting, multiple calligraphy machines exist which were built to replicate some aspects of human artistic writing; however, most projects are limited to a specific style of handwriting, often Chinese calligraphy. We propose a two-stage pipeline that allows industrial robots to write text in arbitrary typefaces and scripts using paintbrushes. In the first stage, we extract a set of strokes from character glyphs which are similar to how humans choose strokes during writing. In the second stage, we generate corresponding brush trajectories by applying a brush model to the extracted strokes. Our brush model computes the required brush pressure to achieve the given stroke width while also accounting for brush lag. We also present a method to automatically measure the parameters needed to predict brush lag by painting and recording calibration patterns. Our method generates trajectories for text in any given typeface, which, when executed by a robotic arm, results in legible written text. We can render most writing systems, excluding emoji and ligatures, in which arbitrary texts can be specified to write.},

keywords = {},

pubstate = {published},

tppubtype = {article}

}

@article{2022,

title = {What Is It about Art? A Discussion on Art.Intelligence.Machine.},

author = {Frederic Fol Leymarie and Seymour Simmons},

url = {http://dx.doi.org/10.3390/arts11050100},

doi = {10.3390/arts11050100},

issn = {2076-0752},

year  = {2022},

date = {2022-10-01},

journal = {Arts},

volume = {11},

number = {5},

pages = {100},

publisher = {MDPI AG},

keywords = {},

pubstate = {published},

tppubtype = {article}

}

@article{10.1145/3505246,

title = {StrokeStyles: Stroke-Based Segmentation and Stylization of Fonts},

author = {Daniel Berio and Frederic Fol Leymarie and Paul Asente and Jose Echevarria},

url = {https://doi.org/10.1145/3505246},

doi = {10.1145/3505246},

issn = {0730-0301},

year  = {2022},

date = {2022-04-01},

journal = {ACM Trans. Graph.},

volume = {41},

number = {3},

publisher = {Association for Computing Machinery},

address = {New York, NY, USA},

abstract = {We develop a method to automatically segment a font’s glyphs into a set of overlapping and intersecting strokes with the aim of generating artistic stylizations. The segmentation method relies on a geometric analysis of the glyph’s outline, its interior, and the surrounding areas and is grounded in perceptually informed principles and measures. Our method does not require training data or templates and applies to glyphs in a large variety of input languages, writing systems, and styles. It uses the medial axis, curvilinear shape features that specify convex and concave outline parts, links that connect concavities, and seven junction types. We show that the resulting decomposition in strokes can be used to create variations, stylizations, and animations in different artistic or design-oriented styles while remaining recognizably similar to the input font.},

keywords = {},

pubstate = {published},

tppubtype = {article}

}

@article{moruzzi2022perceptions,

title = {Perceptions of Creativity in Artistic and Scientific Processes},

author = { Caterina Moruzzi},

doi = {https://doi.org/10.24840/xCoAx_2022_5},

year  = {2022},

date = {2022-01-01},

urldate = {2022-01-01},

journal = {Proceedings of the 10th Conference on Computation, Communication, Aesthetics & X},

pages = {11-25},

abstract = {This paper presents the results of a factorial survey research on perceptions of artistic and scientific creativity in humans and AI. A general reluctance at attributing creativity to artificial systems is well-documented in the literature on the theme. Aim of this survey is to test whether this reluctance is equally strong when participants evaluate scenarios where human and artificial agents are involved in processes of scientific discovery and scenarios where they are engaged in artistic creation processes. The starting hypothesis of the study is that participants should be less hesitant at attributing creativity to artificial agents when the latter engage in scientific discovery processes. Findings, however, disconfirm this assumption, showing that participants attribute significantly less creativity to artificial actors than to human ones, and even more so when they are involved in scientific processes.},

keywords = {},

pubstate = {published},

tppubtype = {article}

}

@article{moruzzi2022creative,

title = {Creative Agents: Rethinking Agency and Creativity in Human and Artificial Systems},

author = {Caterina Moruzzi},

year  = {2022},

date = {2022-01-01},

urldate = {2022-01-01},

journal = {Journal of Aesthetics and Phenomenology},

volume = {9},

number = {2},

pages = {245–268},

publisher = {Taylor & Francis},

abstract = {While the notions of creativity and agency have each received considerable attention in the literature on computational systems, the connections between these two concepts have rarely been addressed. In this paper, I contribute to this debate by discussing the results of an online questionnaire aimed at testing the interactions between the attribution of agency and creativity to human and artificial subjects. Findings of the study indicate that the ascription of agentive capabilities to a subject is positively correlated to the level of creativity attributed to the process they perform. This result supports the claim that a better understanding of the steps needed in order to achieve agency can help also shed light on the development of creativity in computational systems. I conclude the paper by suggesting that the technological revolution of the last decades is compelling us to re-think concepts such as authorship, ownership, agency, and creativity that, so far, have been typically attributed to humans only, and to consider how these notions are being transformed in dynamics of interaction between humans and machines.},

keywords = {},

pubstate = {published},

tppubtype = {article}

}

@article{https://doi.org/10.1111/bjop.12527,

title = {A dot that went for a walk: People prefer lines drawn with human-like kinematics},

author = {Rebecca Chamberlain and Daniel Berio and Veronika Mayer and Kirren Chana and Frederic Fol Leymarie and Guido Orgs},

url = {https://bpspsychub.onlinelibrary.wiley.com/doi/abs/10.1111/bjop.12527},

doi = {https://doi.org/10.1111/bjop.12527},

year  = {2022},

date = {2022-01-01},

journal = {British Journal of Psychology},

volume = {113},

number = {1},

pages = {105-130},

keywords = {},

pubstate = {published},

tppubtype = {article}

}

@article{10.1162/leon_a_02073,

title = {A Machine Learning Application Based on Giorgio Morandi Still-Life Paintings to Assist Artists in the Choice of 3D Compositions},

author = {Guido Salimbeni and Frederic Fol Leymarie and William Latham},

url = {https://doi.org/10.1162/leon_a_02073},

doi = {10.1162/leon_a_02073},

issn = {0024-094X},

year  = {2022},

date = {2022-01-01},

journal = {Leonardo},

volume = {55},

number = {1},

pages = {57-61},

keywords = {},

pubstate = {published},

tppubtype = {article}

}

@article{e24010028,

title = {Network Bending: Expressive Manipulation of Generative Models in Multiple Domains},

author = {Terence Broad and Frederic Fol Leymarie and Mick Grierson},

url = {https://www.mdpi.com/1099-4300/24/1/28},

doi = {10.3390/e24010028},

issn = {1099-4300},

year  = {2022},

date = {2022-01-01},

journal = {Entropy},

volume = {24},

number = {1},

abstract = {This paper presents the network bending framework, a new approach for manipulating and interacting with deep generative models. We present a comprehensive set of deterministic transformations that can be inserted as distinct layers into the computational graph of a trained generative neural network and applied during inference. In addition, we present a novel algorithm for analysing the deep generative model and clustering features based on their spatial activation maps. This allows features to be grouped together based on spatial similarity in an unsupervised fashion. This results in the meaningful manipulation of sets of features that correspond to the generation of a broad array of semantically significant features of the generated results. We outline this framework, demonstrating our results on deep generative models for both image and audio domains. We show how it allows for the direct manipulation of semantically meaningful aspects of the generative process as well as allowing for a broad range of expressive outcomes.},

keywords = {},

pubstate = {published},

tppubtype = {article}

}

@article{arts11040077,

title = {Region-Based Approaches in Robotic Painting},

author = {Jörg Marvin Gülzow and Oliver Deussen},

url = {https://www.mdpi.com/2076-0752/11/4/77},

doi = {10.3390/arts11040077},

issn = {2076-0752},

year  = {2022},

date = {2022-01-01},

journal = {Arts},

volume = {11},

number = {4},

abstract = {An important aspect of robotic painting is replicating human painting techniques on machines, in order to automatically produce artwork or to interact with a human painter. Usually, painterly rendering techniques are transferred to the machine, and strokes are used as the basic building block of an image, as they can easily be mapped to the robot. In contrast, we propose to consider regions as a basic primitive to achieve more human-like results and to make the painting process more modular. We analyze the works of Kadinsky, Mondrian, Delaunay, and van Gogh to show the basis of region-based techniques in the real world and then transfer them to an automatic context. We introduce different types of region primitives and show procedures for how to realize them on our painting machine e-David, capable of painting with visual feedback. Finally, we present machine-created artwork by painting automatically generated sets of shapes in the styles of various artists.},

keywords = {},

pubstate = {published},

tppubtype = {article}

}

@article{doi:10.1177/1357034X211028638,

title = {Interview with Samantha Frost on ‘The Attentive Body’: Epigenetic Processes and Self-formative Subjectivity},

author = {Tomoko Tamari},

url = {https://doi.org/10.1177/1357034X211028638},

doi = {10.1177/1357034X211028638},

year  = {2021},

date = {2021-01-01},

journal = {Body & Society},

volume = {27},

number = {3},

pages = {87-101},

abstract = {The interview is a follow-up from Samantha Frost’s article, ‘The Attentive Body’, in Body & Society 26(4). Tomoko Tamari invites Frost to explore her interest in ‘biocultural creatures’, with its focus on ‘bodies’ responsive self-transformation’ in epigenetic processes, and unfolds Peirce’s account of the index for understanding meaning-making in biological processes. Tamari also introduces Katherine Hayles’s notion of ‘cognitive nonconscious’ to raise the question of the possible theoretical and mechanical similarities/discrepancies between epigenetic processes in organisms and the meaning-making process in computational systems. Drawing on Jacob von Uexkull’s notion of ‘umwelt’ and introducing Yoshimi Kawade’s remarks on a living being’s subjective orientation in environments, a further question about ‘intention’ and ‘subjectivity’ enables Frost to further unpack her notion of ‘the attentive self’ and discuss its relation to ‘intentionality’ and ‘referentiality’ in epigenetic processes. Finally, Samantha Frost mentions her current projects on the connection between ‘attention-as-responsive-self-transformation’ and ‘mode-of-living-as-form-of-life’.},

keywords = {},

pubstate = {published},

tppubtype = {article}

}

@article{moruzzi2021measuring,

title = {Measuring Creativity: An Account of Natural and Artificial Creativity.},

author = {Caterina Moruzzi },

doi = {https://doi.org/10.1007/s13194-020-00313-w},

year  = {2020},

date = {2020-10-02},

urldate = {2020-10-02},

journal = {European Journal for Philosophy of Science},

volume = {11},

number = {1},

pages = {1},

abstract = {Despite the recent upsurge of interest in the investigation of creativity, the question of how to measure creativity is arguably underdiscussed. The aim of this paper is to address this gap, proposing a multidimensional account of creativity which identifies problem-solving, evaluation, and naivety as measurable features that are common among creative processes. The benefits that result from the adoption of this model are twofold: integrating discussions on creativity in various domains and offering the tools to assess creativity across systems of different kinds. By situating creativity within this framework, I aim to contribute to a non-anthropocentric, more comprehensive understanding of the notion, and to debates on natural and artificial creativity.},

keywords = {},

pubstate = {published},

tppubtype = {article}

}

@article{moruzzi2020artificial,

title = {Artificial creativity and general intelligence},

author = { Caterina Moruzzi},

doi = {https://doi.org/10.34632/jsta.2020.9481},

year  = {2020},

date = {2020-01-01},

urldate = {2020-01-01},

volume = {12},

number = {3},

pages = {84-99},

abstract = {It is hard to deny that the notions of creativity and intelligence are

inherently connected. But what does this correlation amount to? Is

creativity a necessary desideratum of intelligence? On the other hand,

does the fact of being intelligent necessarily imply being creative as well?

The aim of this paper is to explore these questions and to contribute to

the discussion regarding the connections between the notions of creativity

and intelligence. In order to do so, I draw on the results obtained from

a study on the public perceptions and attitudes in relation to the use

of AI in the creative sector conducted at the University of Nottingham.

Through this discussion I aim to test the hypothesis that the key features

of creativity correspond to aspects that are essential for the realization

of Artificial General Intelligence, e.g. flexibility, domain knowledge, and

common-sense. After having illustrated the parallels between the two

concepts, I contend that while creativity is a crucial component of general

intelligence, the constituents needed to build an AGI may not be sufficient

to design creative artificial systems. I close the paper by tentatively

suggesting how the motivations behind the discontent expressed by the

participants against creative AI can be explained through the uncanny

valley phenomenon.},

keywords = {},

pubstate = {published},

tppubtype = {article}

}

@article{app10103396,

title = {Recent Developments Regarding Painting Robots for Research in Automatic Painting, Artificial Creativity, and Machine Learning},

author = {Jörg Marvin Gülzow and Patrick Paetzold and Oliver Deussen},

url = {https://www.mdpi.com/2076-3417/10/10/3396},

doi = {10.3390/app10103396},

issn = {2076-3417},

year  = {2020},

date = {2020-01-01},

journal = {Applied Sciences},

volume = {10},

number = {10},

abstract = {E-David (Electronic Drawing Apparatus for Vivid Image Display) is a system for controlling a variety of painting machines in order to create robotic paintings. This article summarizes the hardware set-up used for painting, along with recent developments, lessons learned from past painting machines, as well as plans for new approaches. We want to apply e-David as a platform for research towards improving automatic painting and to explore machine creativity. We present different painting machines, from small low-cost plotters to large industrial robots, and discuss the benefits and limitations of each type of platform and present their applicability to different tasks within the domain of robotic painting and artificial creativity research. A unified control interface with a scripting language allows users a simplified usage of different e-David-like machines. Furthermore, we present our system for automated stroke experimentation and recording, which is an advance towards allowing the machine to autonomously learn about brush dynamics. Finally, we also show how e-David can be used by artists “in the field” for different exhibitions.},

keywords = {},

pubstate = {published},

tppubtype = {article}

}

@article{chamberlain2018putting,

title = {Putting the art in artificial: Aesthetic responses to computer-generated art.},

author = {Rebecca Chamberlain and Caitlin Mullin and Bram Scheerlinck and Johan Wagemans},

url = {https://research.gold.ac.uk/id/eprint/20443/1/ComputerGeneratedArt.pdf},

year  = {2018},

date = {2018-01-01},

urldate = {2018-01-01},

journal = {Psychology of Aesthetics, Creativity, and the Arts},

volume = {12},

number = {2},

pages = {177},

publisher = {Educational Publishing Foundation},

keywords = {},

pubstate = {published},

tppubtype = {article}

}

@article{arts7040084,

title = {Self-Improving Robotic Brushstroke Replication},

author = {Jörg Marvin Gülzow and Liat Grayver and Oliver Deussen},

url = {https://www.mdpi.com/2076-0752/7/4/84},

doi = {10.3390/arts7040084},

issn = {2076-0752},

year  = {2018},

date = {2018-01-01},

journal = {Arts},

volume = {7},

number = {4},

abstract = {Painting robots, like e-David, are currently unable to create precise strokes in their paintings. We present a method to analyse given brushstrokes and extract their trajectory and width using a brush behaviour model and photographs of strokes painted by humans. Within the process, the robot experiments autonomously with different brush trajectories to improve the reproduction results, which are precise within a few millimetres for strokes up to 100 millimetres length. The method can be generalised to other robotic tasks with imprecise tools and visible results, like polishing or milling.},

keywords = {},

pubstate = {published},

tppubtype = {article}

}

@article{TRESSET2013348,

title = {Portrait drawing by Paul the robot},

author = {Patrick Tresset and Frederic Fol Leymarie},

url = {https://www.sciencedirect.com/science/article/pii/S0097849313000149},

doi = {https://doi.org/10.1016/j.cag.2013.01.012},

issn = {0097-8493},

year  = {2013},

date = {2013-01-01},

journal = {Computers & Graphics},

volume = {37},

number = {5},

pages = {348-363},

abstract = {We describe Paul, a robotic installation that produces observational face drawings of people. Paul is a naive drawer: it does not have highlevel knowledge of the structures constitutive of the human face (such as the mouth, nose, eyes) nor the capability of learning expertise based on experience as a human would. However, Paul is able to draw using the equivalent of an artist's stylistic signature based on a number of processes mimicking drawing skills and technique, which together form a drawing cycle. Furthermore, we present here our first efforts in implementing two different versions of visual feedback to permit the robot to iteratively augment and improve a drawing which initially is built from a process of salient lines recovery. The first form of visual feedback we study we refer to as computational as it involves a purely internal (memory-based) representation of regions to render via shading by the robot. The second version we call physical as it involves the use of a camera as an ‘eye’ taking new snapshots of the artefact in progress. This is then analysed to take decisions on where and how to render shading next. A main point we emphasise in this work is the issue of embodiment of graphical systems, in our case in a robotic platform. We present our arguments in favour of such a position for the graphics community to reflect upon. Finally, we emphasise that the drawings produced by Paul have been considered of interest by fine art professionals in recent international art fairs and exhibitions, as well as by the public at large. One drawing is now in the Victoria and Albert museum collection. We identify a number of factors that may account for such perceived qualities of the produced drawings.},

keywords = {},

pubstate = {published},

tppubtype = {article}

}