This document provides an introduction to neuroaesthetics, which examines the neural processes underlying aesthetic experiences such as art appreciation. It discusses several key topics:
1. Definitions of neuroaesthetics and debates around whether it should encompass the study of art production and appreciation or have a narrower focus.
2. Evidence that aesthetic priming, such as images related to painting techniques, can influence aesthetic preferences for art, supporting the role of covert simulation in appreciation.
3. Research on the timing of neural responses to artistic style and content, finding that style information is available 40-94ms later than content.
4. Studies using ERPs to analyze the time course of visual, cognitive
Designing IA for AI - Information Architecture Conference 2024
Artificial Intelligence for Visual Arts
1.
2. Introduction to the format
●
Shallow introduction for the use of artificial intelligence in visual arts both in computational
aesthetics and fine art analysis in dense 'mind map' format.
– Best to be read from a tablet (or similar device with easy zoom in/out), do not work very well for project
despite the slide format.
– Intention is to provide 'leads' to many different fields making it easier for the interested reader to search more
knowledge on the relevant topics.
NEUROAESTHETIC ANALYSIS
●
Separation of aesthetics and art, defining aesthetically pleasing or novel works is a lot easier than defining 'good
and original art' per se.
– Ideally, we would like to combine the computational aesthetics and neuroaesthetic foundation into a single generative
and predictive framework.
ART MARKET ANALYTICS
●
In theory, the predictive analytics model of art market is a multidisciplinary collaboration with neuroaesthetics,
computational aesthetics, artificial intelligence (~deep learning), art history and quant trading.
●
In practice however the predicting the prices of the whole market (and even more a single artist) is extremely challenging:
– Art market is inefficient with price formation being opaque.
– Pleasantness or popularity of visual art can be predicted rather well with current deep learning techniques that would
not correlate with the value of those works necessarily.
– Single institutions (auction house,art critics, powerful collectors) can drive the prices of given artist quickly either
down or up (e.g. Charles Saatchi)
3. Motivation
Building a generative cross-disciplinary deep learning model for predicting
the visual appeal and originality, and art market pricing
NEUROAESTHETICS
ART MARKET
DOMAIN KNOWLEDGE
4. NeuroAesthetics background
How does aesthetic experience rise?
Summary:
Predict what visual characteristics people response favorably.
• Whether there are subgroups such as layman vs. art
connoisseurs?
• What visual characteristcs have corresponded to high end art
market valuations during different periods?
• What sort of quantitative change have happened from transition
from period to another?
• Could we come come up with generative artificial intelligence
algorithms that would incorporate the domain knowledge of
neuroaesthetics in creating art?
5. Neuroaesthetics introduction #1
https://www.ucl.ac.uk/cdb/research/zeki, Professor Semir Zeki,
“Pioneer of neuroaesthetics”
https://youtu.be/NlzanAw0RP4
https://neuroaesthetics.net/papers/visual-arts/
the-scientist.com
But what in the brain triggers aesthetic experiences? And how does
knowledge of basic brain mechanisms inform our understanding of these
experiences? These questions are at the heart of an emerging discipline
dedicated to exploring the neural processes underlying our appreciation and
production of beautiful objects and artwork, experiences that include
perception, interpretation, emotion, and action. This new field represents a
convergence of neuroscience and empirical aesthetics—the study of
aesthetics rooted in observation—and is dubbed neuroaesthetics, a term
coined in the 1990s by vision neuroscientist Semir Zeki of University College
London.
Neuroaesthetics is both descriptive and experimental, with qualitative
observations and quantitative tests of hypotheses, aimed at advancing our
understanding of how humans process beauty and art. While the field is still
young, interest is growing, as evidenced by several recent books on the
topic. Moreover, recent workshops such as “Pain and Pleasure: The Art and
Science of Body Representation,” held in Venice last November, and “Visual
Arts and the Brain,” held at the Royal Society of Medicine in London the
following month, demonstrate the international scope of this discipline as it
addresses various aesthetic experiences, and their underlying neural
correlates, in health and in disease.
Anjan Chatterjee is the author of The Aesthetic Brain: How We Evolved to Desire Beauty and Enjoy Art
and coeditor of Neuroethics in Practice: Mind, Medicine, and Society and The Roots of Cognitive
Neuroscience: Behavioral Neurology and Neuropsychology.
6. Neuroaesthetics introduction #2
http://dx.doi.org/10.1016/j.bandc.2011.01.009
http://dx.doi.org/10.1093/brain/awv163
Today, neuroaesthetics is a rather heterogeneous research area: scientists have
entered the field with different backgrounds, interests and questions in mind.
Hence, there is not necessarily a consensus on what the important questions
are or about how best to produce answers.
One of the main issues raised at the Conference was the definition and scope of
the field. Neuroaesthetics is often conceived as the study of the neural basis of
the production and appreciation of artworks
However, Brown and Dissanayake (2009) argued that because art goes beyond
aesthetic concerns, this definition is too broad in that it attempts to account for
the biological underpinnings of artistic behavior, which includes a number of
cognitive and affective mechanisms that have no aesthetic relevance. Hence,
they contend that in addition to neuroaesthetics, a field of neuroartsology is
required.
In contrast to this view, authors such as Skov and Vartanian (2009a) have used
the term neuroaesthetics in a rather more general way to encompass the study
of the biological roots of the variety of psychological and neural processes
involved in the creation andperception of artistic and non-artistic objects. In this
sense, neuroaesthetics is close to what
Fitch, von Graevenitz, and Nicolas (2009) have defined as bioaesthetics.
One of the future directions neuroaesthetics must explore is the time course of
brain activation related with aesthetic experiences. Researchers need to move
beyond mere localization of brain areas engaged in such experiences to produce
a dynamic view of neural processes. This is almost self-evidently true in the
case of music and dance but also applies to the appreciation of visual art and
architecture. Another important issue for future research is the identification of
genuine modality-independent processes – distinct from modality-specific
ones – involved in artistic and aesthetic appreciation involving different sensory
modalities, such as music, dance, painting, and so on.
Book review of AN INTRODUCTION TO NEUROAESTHETICS. The Neuroscientific Approach to Aesthetic
Experience, Artistic Creativity, and Arts Appreciation Edited by Jon O. Lauring, 2014. Copenhagen: Museum
Tusculanum Press ISBN: 978 87 635 4140 4
Definitions matter when investigating the neurological basis for aesthetics and beauty, as these
qualities are subjective and arbitrary—witness that the artist Jack Vettriano sold over 10 million
copies of his painting The Singing Butler, and the picture itself for £744 500 in 2004, yet it was
rejected by the Royal Academy when it was entered for one of its annual summer exhibitions,
and many art experts have reportedly judged the artist’s work unfavourably (Malvern, 2015).
Assuming that individuals’ brain circuitry might be broadly similar, how could one explain in
neurological terms the differences in aesthetic appreciation? Is it a question of
experience and memory, or is it attributable to ‘… the personal, social, cultural, and educational
history that have shaped the beholder’s personality’ (p. 116), prestige and monetary factors (pp.
138–9), or even prejudice?
Having briefly considered some of the many aspects pertaining to neuroaesthetics, it is only fair
to also acknowledge some people’s scepticism about the whole subject and adopt the role of
devil’s advocate by asking: is this subject valid and worthy of study? Massey’s and Seeley’s
criticisms have been referred to above, and Casati and Pignocchi (2007) have dismissed the role
of mirror and canonical neurons in the aesthetic response. But the most critical view is that
held by the neuro-philosopher Raymond Tallis:
‘It is perfectly obvious why we might expect neuroaesthetics to remain a sterile as well as an
almost comically simplistic exercise, even more misguided than trying to explain the genius of a
ballet dancer using electromyography. Paintings are treated as mere isolated stimuli or sets of
stimuli … Works of art are not merely sources of stimuli that act on bits of the brain’ (Tallis, 2008).
Could not the same criticisms apply to music, literature, and all the other creative arts?
https://newhumanist.org.uk/articles/2172/neurotrash
7. Neuroaesthetics introduction #3
DOI: 10.1177/1745691615621274
Leder, Belke, Oeberst, and Augustin’s (2004) model of aesthetic experience of art.
http://dx.doi.org/10.1037/a0031842
http://dx.doi.org/10.3389/fnhum.2014.00891
8. Neuroaesthetics introduction #4: Priming
https://dx.doi.org/10.3389%2Ffnhum.2014.00391
http://dx.doi.org/10.1007/s00221-015-4395-5
http://dx.doi.org/10.1371/journal.pone.008015
4
(A) Images of a gloved-hand holding a paintbrush were used as supraliminal priming before the display of each
pointillist-style painting. The images consisted of either a precision or a power grip, or of a rested palm down
hand and they created three conditions. Compatible (precision grip) or Incompatible (power grip) with the
drawing of pointillist-style paintings. The palm down image served as Control. (B) The preference expressed
when the paintings were preceded by priming images activating motor programs Compatible with the
production of pointillist-style brushstrokes was higher than that expressed for the Incompatible and the Control
conditions.
In conclusion, we here provide empirical evidence that, beyond other factors
such as upbringing, historical context and nature of the artistic stimuli, covert
painting simulation may influence affective responses to art (
Freedberg and Gallese, 2007). Obtaining a better understanding of the
contribution of action simulation in affective states is likely to shed light not
just on how the brain encodes affective stimuli but also may enrich our
perspective on the neural mechanisms involved in some social and
communicative deficits associated with action simulation, such as autism
spectrum disorder (Oberman and Ramachandran, 2007).
We show that action priming, when congruent with the artist's painting
style, enhanced aesthetic preference. These results support the hypothesis
that involuntary covert painting simulation contributes to aesthetic
appreciation during passive observation of artwork.
Studies using affective priming procedures demonstrate, for example, that inducing a conscious negative
emotional state biases the perception of abstract stimuli towards the sublime. Moreover, subliminal happy
facial expressions have a positive impact on the aesthetic evaluation of abstract art. Little is known about
how emotion influences aesthetic perception of non-abstract, representational stimuli,
especially those that are particularly relevant for social behaviour, like human bodies. Here, we explore
whether the subliminal presentation of emotionally charged visual primes modulates the explicit
subjective aesthetic judgment of body images. Using a forward/backward masking procedure, we
presented subliminally positive and negative, arousal-matched, emotional or neutral primes and measured
their effect on the explicit evaluation of perceived beauty (high vs low) and emotion (positive vs negative)
evoked by abstract and body images. We found that negative primes increased subjective
aesthetic evaluations of target bodies or abstract images in comparison with positive primes. The
study expands previous research by showing the effect of subliminal negative emotions on the
subjective aesthetic evaluation not only of abstract but also of body images.
9. Neuroaesthetics introduction #5: ERP Timing
http://dx.doi.org/10.1016/j.neuropsychologia.2011.03.038, Cited by 26 http://dx.doi.org/10.3389/fnhum.2015.00525,
Cited by 3
Twenty-two participants viewed pictures that systematically varied in style and content
and conducted a combined go/nogo dual choice task. The dependent variables of
interest were the Lateralised Readiness Potential (LRP) and the N200 effect.
Analyses of both measures support the notion that in the processing of art style
follows content, with style-related information being available at around 224ms or
between 40 and 94ms later than content-related information. The paradigm used here
offers a promising approach to further explore the time course of art perception, thus
helping to unravel the perceptual and cognitive processes that underlie the
phenomenon of art and the fascination it exerts.
Important issue concerns the participants’ background knowledge. Many
articles on art perception and aesthetics (e.g., Augustin & Leder, 2006; Belke et al., 2010;
Cupchik & Gebotys, 1988; Hekkert&vanWieringen, 1996; Leder et al., 2004; Wiesmann & Ishai, 2010)
assume a high relevance of art-related expertise for the classification and
evaluation of artworks, especially in the context of the processing of style. In order to
find out whether the time course of art perception proposed here is a relatively
general phenomenon or to what extent it depends on expertise, future studies with
carefully selected samples of art-experts and non-experts are needed.
Grand average ERPs (n = 22) for go and nogo trials for the two dual choice conditions hand = content, go/nogo =
style and hand = style, go/nogo = content at electrode sites Pz (Fz, Cz cropped). The right column illustrates the
results regarding the N200 effect: Each graph plots the difference waves nogo–go for the two dual choice conditions.
The x-axes represent the time relative to picture onset, the y-axes plot activation in μV, with negative voltage plotted
up.
To study the time course of visual, cognitive and emotional processes in response to visual art
we investigated the event-related potentials (ERPs) elicited whilst viewing and rating the visceral affect
of three categories of visual art. Two groups, artists and non-artists viewed representational,
abstract and indeterminate 20th century art. Early components, particularly the N1, related to attention
and effort, and the P2, linked to higher order visual processing, was enhanced for artists when
compared to non-artists. This effect was present for all types of art, but further enhanced for abstract
art (AA), which was rated as having lowest visceral affect by the non-artists. The later, slow wave
processes (500–1000 ms), associated with arousal and sustained attention, also show clear
differences between the two groups in response to both type of art and visceral affect. AA increased
arousal and sustained attention in artists, whilst it decreased in non-artists. These results suggest that
aesthetic response to visual art is affected by both expertise and semantic content.
Research into viewing habits of gallery
visitors suggests that the average time spent
contemplating art in galleries is 30 s (
Locher et al., 2007). Although differences between
schools of art have been identified in 1 ms (
Bachmann and Vipper, 1983, as cited in
Augustin et al., 2008), and here we demonstrated
differences in visual and visceral responses in
presentation times of less than 1500 ms, perhaps
these differences are not specific to art, but are
simply in response to visual stimuli. Most art is
created to be contemplated, to be thought
provoking, and to engage. In order to ensure
responses are to art longer presentation times could
be employed, in art galleries. Finally, the impact of
expertise could be further explored. Does expertise
impact on visual and affective processes more
generally? Do art experts see faster,
differently, more?
Grand average topographic scalp maps showing ERP
components for the late positive component, (LPP, at
1000 ms), for abstract art (AA), representational art (RA)
and Indeterminate art (IA), for two groups, artists (A) and
non-artist (NA) for 12 electrodes.
10. 'Classical aesthetics' resistance to 'neuropressure'
Observations on the
Feeling of the Beautiful
and Sublime
by Immanuel Kant
Aesthetics: An
Introduction to the
Philosophy of Art
by Anne Sheppard
philosophytalk.org/community/blog/laura-maguire/2016/08
http://dx.doi.org/10.1111/jaac.12283
I offer a Darwinian perspective on the existence of aesthetic interests, tastes,
preferences, and productions. It is distinguished from the approaches of Denis Dutton
and Geoffrey Miller, drawing instead on Richard O. Prum's notion of biotic artworlds.
The relevance of neuroaesthetics to the philosophy of art is defended.
http://dx.doi.org/10.1057/9780230369580_13
How the “Continentals”
Internalized Their Oppressors
Or in other words, by means of The Sokal Hoax, the
Continentals were professionally hyper-disciplined by their
Analytic oppressors.
7. Neuro WtF?
Duke and Vanderbilt universities now have neuroscience
centers with specialties in humanities hybrids, from
“neurolaw” onward: Duke has a Neurohumanities
Research Group and even a neurohumanities abroad
program. The money is serious as well. Semir Zeki, a
neuroaesthetics specialist—that is, neuroscience applied
to the study of visual art—was the recipient of a £1 million
grant in the United Kingdom.
Deena Skolnick Weisberg, a Temple University
postdoctoral fellow in psychology, wrote a 2008 paper
titled
“The Seductive Allure of Neuroscience Explanations,” in
which she argued that the language of neuroscience
affected nonexperts’ judgment, impressing them so much
that they became convinced that illogical explanations
actually made sense. Similarly, combining neuroscience
with, say, the study of art nowadays can seem to offer an
instant sheen of credibility.
arcade.stanford.edu
againstprofphil.org
11. Neuroaesthetics Possible problems
http://dx.doi.org/10.1038/nature.2013.12640
http://dx.doi.org/10.1371/journal.pbio.1001504
Right or wrong? If anyone is going to pursue neuroaesthetics, I’d be glad for it to be
Zeki, who has a deep and sincere appreciation of art, and an awareness of the limits of a
scientific approach to the way we experience it. But some of the pitfalls of
neuroaesthetics are perceptively expressed by neuroscientist Bevil Conway of Wellesley
College and musicologist Alexander Rehding of Harvard University in Cambridge, both in
Massachusetts, in an essay that appeared this week in PLoS Biology.
They point out that “it is an open question whether an analysis of artworks, no matter
how celebrated, will yield universal principles of beauty” and that “rational reductionist
approaches to the neural basis for beauty ... may well distill out the very thing one wants
to understand”.
For one thing, to suggest that the human brain responds in a particular way to art risks
creating criteria of right or wrong, either in the art itself or in individual reactions to it.
Although it is a risk that most researchers are likely to recognize, experience suggests that
scientists studying art find it hard to resist drawing up rules for critical judgements. The
chemist and Nobel laureate Wilhelm Ostwald, a competent amateur painter, devised an
influential theory of colour in the early twentieth century that led him to declare that
Titian had once used the ‘wrong’ blue. Paul Klee, whose intuitive handling of colour was
impeccable, spoke for many artists in his response to such hubris:
“That which most artists have in common, an aversion to colour as a science, became
understandable to me when, a short time ago, I read Ostwald’s theory of colours ...
Scientists often find art to be childish, but in this case, the position is inverted ... To
hold that the possibility of creating harmony using a tone of equal value should become
a general rule means renouncing the wealth of the soul. Thanks but no thanks.”
Even if neuroaestheticists refrain from making similar value judgements, they are already
close to falling prey to one. Conway and Rehding discuss this field primarily as an attempt
to understand how the brain responds to beauty. As they point out, beauty is not a
scientific concept — so it is not clear which questions neuroaesthetics is even
examining.
Immanuel Kant is a preferred philosopher among neuroaestheticians, no doubt because of his
towering stature in the history of Western thought. He pursued a universalist approach to beauty, an
appealing concept for neuroscientists because it suggests a discrete neural basis. But Kant's concept
of beauty has been severely criticized in light of the prevailing pluralism of artistic styles. To make
matters more complicated, there is no consensus on the nature of beauty. Kant's understanding of
beauty was predicated on an attitude of “disinterested contemplation”[2], whereas Friedrich
Nietzsche roundly dismissed this notion and underlined the impact of sensual attraction [3]. For the
poet John Keats, beauty equaled truth [4], while Stendhal, the French novelist, characterized beauty
as the “promise of happiness” [5]. More recently, Elaine Scarry described beauty as an urge to
repeat [6]. While each of these theories is respected, not one is universally accepted. Partly this
diversity of opinions is connected to the different functions that beauty holds within various
philosophical systems, being sometimes viewed in connection with epistemology or with ethics. One
goal of neuroaesthetics is to get to the bottom of the problem of artistic beauty. How can this be
accomplished?
It may be no coincidence that the art relates to the culture in which each artist were raised. One
potential danger in aesthetic projects is to universalize one's subjective convictions and assume that
an experience of beauty is common to all. Projecting from individual subjective experience is
deceptive, for there is ample evidence that notions of beauty vary between cultures and are mutable
even within a culture—just think of fast-changing trends in fashion. Moreover, the equation (art =
beauty) rests on shaky ground. Throughout history, artists have created deeply moving artwork that
is emphatically not beautiful; Goya's Saturn Devouring One of His Sons provides a famous historical
example.
There may well be a “beauty instinct” implemented by dedicated neural machinery capable of
producing a diversity of beauty reactions, much as there is language circuitry that can support a
multitude of languages (and other operations). A need to experience beauty may be universal, but
the manifestation of what constitutes beauty certainly is not. On the one hand, a neuroaesthetics
that extrapolates from an analysis of a few great works, or one that generalizes from a single specific
instance of beauty, runs the risk of missing the mark. On the other, a neuroaesthetics comprising
entirely subjectivist accounts may lose sight of what is specific to encounters with art.
Neuroaesthetics has a great deal to offer the scientific community and general public. Its progress in
uncovering a beauty instinct, if it exists, may be accelerated if the field were to abandon a pursuit of
beauty per se and focus instead on uncovering the relevant mechanisms of decision making and
reward and the basis for subjective preferences, much as Gustav Fechner counseled. This would mark
a return to a pursuit of the mechanisms underlying sensory knowledge: the original conception of
aesthetics.
12. Neuroaesthetic 'engine' Universality?
http://journal.frontiersin.org/researchtopic/2967
http://dx.doi.org/10.1111/j.1469-7580.2009.01164.
x
http://dx.doi.org/10.3389/fpsyg.2016.00750
Although the reasons individuals have specific stable aesthetic preferences—for
example, for abstract art or for classical music—are often studied (e.g.,
Furnham and Walker, 2001), there is a growing stream of research (e.g., Nodine et al., 1993;
Chamorro-Premuzic and Furnham, 2004; Axelsson, 2007; Kozbelt and Seeley, 2007; Silvia, 2007; Myszkowski et al., 2014)
that is interested in the various abilities involved when evaluating art: Are we all
equally “armed” to process aesthetic stimuli?
Our aim in this paper is to propose a new direction for this stream of research. While
a typical approach to the study of aesthetic ability consists in measuring single
facets, notably aesthetic sensitivity (e.g. ,Myszkowski et al., 2014), we propose a multi-
content approach. More specifically, mirroring the “g-to-IQ” shift in intelligence
measurement, we want to propose a “T-to-AQ” shift from single-content measures
of “good taste” (“T”) to comprehensive assessments of an “Aesthetic Quotient”
(AQ), which would include other facets of aesthetic ability—like artistic knowledge,
sensitivity to complexity and aesthetic empathy. Rather that questioning
the existence of an AQ, we argue its usefulness, notably in predicting creative
potential and achievement.
We have proposed for this approach the term Aesthetic Quotient (AQ), as a
reference to the “g-to-IQ” shift to comprehensive assessments of intelligence (e.g.,
Wechsler, 2008). We finally explained that psychology and empirical aesthetics
researchers should probably stay clear of philosophical debates on the existence of
aesthetic ability, and rather focus on the accumulating evidence on the usefulness of
AQ components as predictors of creative potential and achievement.
http://dx.doi.org/10.3389/fnhum.2016.00213
13. Psychological models of Art experience
http://dx.doi.org/10.3389/fnhum.2016.00160
Chatterjee model adapted from original visual model in Chatterjee (2004). Original
elements shown in black. Additions not originally included in model shown in blue. If
possible, original wording has been retained or adapted from model author's publications.
Locher model (adapted from Locher, 1996; Locher et al., 2010).
Leder model (adapted from Leder et al., 2004; Leder and Nadal, 2014).
Cupchik model (created by the authors for this paper).Pelowski model (adapted from Pelowski and Akiba, 2011).
Silvia model (created by the authors for this paper).
Missing Elements: Physiology, Health, Negative and Profound Reactions to Art
Regarding inputs, there are also areas for future development. Specific artwork-related aspects such as style are not included in several models (Chatterjee, Silvia,
Pelowski). The same can be said for the artwork's historical context, which was also recently argued to be a key processing input (Bullot and Reber, 2013), but in
the present review only operationalized as one aspect of the background knowledge of viewers (e.g., by Locher and Leder, but see Pelowski and Akiba, 2011). It
also appears that only the models put forward by Leder and Cupchik account for the current psychophysiological and affective state of the viewer. These aspects
should be incorporated into the other models and systematically included when setting up experiments. In addition, while most authors specifically note the
importance of memory components for processing, and often mention this in their written theory, it is often omitted in the models. This begs for integration and
elaboration.
15. 'Traditional correlation' studies Intro #1
http://dx.doi.org/10.1016/j.visres.2010.05.002
(A) Original images (A, D, G, J), their Fourier power spectra (B, E, H, K) and
log–log plots of radially averaged Fourier power versus spatial frequency
(C, F, I, L). In the Fourier spectra (B, E, H, K), the low spatial frequencies
are represented at the center and lighter shades represent higher power. In
the log–log plots (C, F, I, L), straight lines are fitted to binned data points
between 10 and 256 cycles/image
Log–log plots of radially averaged Fourier power
versus spatial frequency. Average curves are
given for the datasets of face photographs (AR
face database; Martinez & Benavente, 1998), for
200 examples from the Groningen natural scene
database (Van Hateren & van der Schaaf, 1998)
and for 306 monochrome art portraits of
Western provenance. Note that the average
curves for the art portraits and natural scenes
have a similar slope and are more shallow than
the average curve for the face photographs (see
also Table 1). Data modified after
Redies, Hänisch et al. (2007). a Standard deviation.
b Number of images in each category.
c Data from the study by Redies, Hasenstein et al. (2007).
d Images from the database of van Hateren and van der Schaaf (1998).
e Data from the study by Redies, Hänisch et al. (2007).
f AR face database of Martinez and Benavente (1998).
A number of studies have examined statistical properties of art beyond the basic image
statistics described above. Here, we give only some examples of such statistics, which relate to
gross properties of composition. A complete survey of composition-level statistics is beyond the
scope of the present review (see, e.g., Tyler, 2007).
Cheyne, Meschino, and Smilek (2009) found consistent variations in the relationships among marks in Paleolithic cave
painting. In this innovative study, statistical relationships between anatomical features in animals and their painted
representation were evaluated. Results suggest that these variations are a form of caricature, and that they therefore stand as
evidence of human perception of typicality and categorization. The authors propose that cave artists were keen observers of
the wildlife they painted (cf. Guthrie, 2005), and that individual variations in composition reflected the deep knowledge of
these animals that contemporary viewers would readily understand. This result relates to the proposal of
Ramachandran and Hirstein (1999), who argued that representational art in many cultures is largely directed towards
identifying and exaggerating distinctive features in the manner of caricature. These authors see art as an aesthetic
manifestation of the “peak-shift effect,” wherein animals (particularly the young) respond more favorably to exaggerated,
counterfeit versions of relevant stimuli than to “natural” versions of those same stimuli.
17. 'Traditional correlation' studies abstract features
http://dx.doi.org/10.3389/fpsyg.2016.00973
By means of a previously established MATLAB algorithm (Amirshahi et al., 2012),
we calculated the following SIPs for every single image from the dataset:
(1) PHOG Self-Similarity. Here, we calculated Self-Similarity using the Pyramid Histogram of
Orientation Gradients (PHOG) method that was introduced by Bosch et al. (2007). The
algorithm is based a comparison of histograms of oriented gradients (HOGs) from the
entire image with HOGs from equal subparts of the image. For a detailed description of
the procedure, see the Appendix in Braun et al. (2014). Self-Similarity, a concept closely
related to scale-invariance and fractality, implies that an object has a structure similar to
its parts. Museum paintings exhibit a relatively high degree of Self-Similarity compared to
other image categories (Amirshahi et al., 2012, 2013; Redies et al., 2012).
(2) HOG Complexity. Berlyne (1974) postulated that an intermediate complexity of stimuli
leads to a higher aesthetic appeal than low or high complexity (Nadal, 2007). Recently,
several studies confirmed the role of complexity in beauty perception (
Jacobsen and Hofel, 2002; Rigau et al., 2008;Forsythe et al., 2011). We defined HOG
Complexity as the sum of the strengths of the oriented gradients in the image as
described by Braun et al. (2014).
(3) Anisotropy. Anisotropy is a measure for the distribution of orientation of gradients within
a particular image. Low Anisotropy implies that the strength of luminance gradients is
uniformly distributed across all orientations; high values indicate that one or a few
orientations are represented more strongly than others in the orientation spectrum.
Previous studies showed that colored artworks show a relatively low degree of Anisotropy
compared to other categories of images (Redies et al., 2012). We calculated Anisotropy as
described by in Braun et al. (2014).
(4) Aspect Ratio. Although there is no evidence for an overall preference of a certain format
of paintings (McManus, 1980;Russell, 2000), we used this measure to investigate whether
it is correlated with the subjective description of images and whether certain groups of
participants preferred a certain aspect ratio over others in abstract artworks. The measure
was obtained by dividing image height by image width.
(5) Color measures. In addition to second-order image statistics, we calculated the three
color measures of the HSV color space (Color Hue, Color Saturation and Color Value),
which have been used in aesthetic quality assessment of images previously (
Datta et al., 2006). Previous studies described a link between color and emotion (
Ou et al., 2004a). We calculated the color measures by means of a MATLAB algorithm.
The HSV values were computed pixel-by-pixel. For each of the three color measures, the
mean across all pixels was taken as the final value.
Examples for test images. (A) Monument im Fruchtland, Paul Klee,
1929; (B) Mystic Suprematism (red cross on black circle), Kazimir
Malevich, 1920–1922; (C) Untitled VIII, Willem de Kooning, 1980(c)
The Willem de Kooning Foundation, New York/VG Bild-Kunst, Bonn,
2016; (D) Stretched Yellow, Lee Krasner, 1955(c) Pollock-Krasner
Foundation/VG Bild-Kunst, Bonn, 2016.
18. 'Traditional correlation' studies Fractality #1
FRACTALITY
●
Beauvois, Michael W. ‘Quantifying Aesthetic Preference And Perceived Complexity For Fractal Melodies’. Music Perception 24, no. 3 (February 2007): 247–
64. doi: http://dx.doi.org/10.1525/mp.2007.24.3.247
●
Goldberger, A L. ‘Fractals and the Birth of Gothic: Reflections on the Biologic Basis of Creativity’. Molecular Psychiatry 1, no. 2 (May 1996): 99–104.
http://www.ncbi.nlm.nih.gov/pubmed/9118332.
●
Jones-Smith, Katherine, and Harsh Mathur. ‘Fractal Analysis: Revisiting Pollock’s Drip Paintings’. Nature 444, no. 7119 (30 November 2006): E9–10. doi:
http://dx.doi.org/10.1038/nature05398.
●
Jones-Smith, Katherine, Harsh Mathur, and Lawrence M. Krauss. ‘Drip Paintings and Fractal Analysis’. Physical Review E 79, no. 4 (30 April 2009): 46111. doi:
http://dx.doi.org/10.1103/PhysRevE.79.046111
●
Joye, Yannick. ‘A Review of the Presence and Use of Fractal Geometry in Architectural Design’. Environment and Planning B: Planning and Design 38, no. 5
(2011): 814 – 828. doi:http://dx.doi.org/10.1068/b36032
●
Redies, Christoph. ‘A Universal Model of Esthetic Perception Based on the Sensory Coding of Natural Stimuli’. Spatial Vision 21 (December 2007): 97–117.
doi: http://dx.doi.org/10.1163/156856807782753886
http://dx.doi.org/10.3389/fnhum.2016.00210
http://dx.doi.org/10.1038/nature05398
The best broken power-law fit to these data
corresponds to slopes of DD = 1.53 and DL = 1.84.
The break occurs at ln(L) 4; the standard deviation of
the data from the fit, , is 0.022. Untitled 5 (top) fulfils
all the criteria used in box-counting authentication
that have been made public: it shows a broken
power-law behaviour with DD < DL and, for a
magnification factor C similar to that used by Taylor et
al., has a 2value in the 'permissible' range of 0.009
< < 0.025.
Methods. All our sketches, including Untitled 5, are
freehand drawings made in Adobe Photoshop using a
14-point Adobe Photoshop 'paintbrush'. The
paintbrush leaves a mark when dragged continuously
across the 'canvas by a computer mouse. Although
not drip paintings, these patterns are human- and not
computer-generated.
a, Top, a middle-third Cantor dust anchor layer (blue), overlaid with a second Cantor dust (red). Half-
blue/half-red bars correspond to the intersection of the dusts; purple corresponds to their union. Centre,
box-counting curves: blue dust (shown in blue), red dust (red), the uncovered part of the blue dust (green)
and the composite (purple). The curvature of the traces indicates that the uncovered portion of the blue
layer and the composite are not true fractals. To highlight the curvature of the composite, the lower graph
shows the difference of the (rigorously linear) blue and purple curves. b, A linear fit to the box-counting
curve of a 100-step gaussian walk has a slope of 1.35, with standard deviation = 0.025.
Methods. For a, the blue dust is obtained by repeatedly dividing segments into three parts and retaining
only the first and third; the red dust is obtained by dividing into nine parts and retaining the first, fifth and
ninth parts. The 'fractal barcode' shows the appearance of the dusts after four iterations. For b, step size is
calculated as 0.09 frame width. Smallest box size, 3 pixels. Sizes range over 1.4 orders of magnitude, with
magnification C = 1.12 (see Fig. 2 for definition).
This suggests that exact fractals are processed differently than
their statistical counterparts. We propose a set of four factors
that influence complexity and preference judgments in fractals
that may extend to other patterns: fractal dimension,
recursion, symmetry and the number of segments in a pattern.
Conceptualizations such as Berlyne’s and Redies’ theories of
aesthetics also provide a suitable framework for interpretation
of our data with respect to the individual differences that we
detect. Future studies that incorporate physiological methods
to measure the human aesthetic response to exact fractal
patterns would further elucidate our responses to such
timeless patterns.
19. 'Traditional correlation' studies Signal analysis
The fractal dimension D of Pollock
paintings plotted against the year in which
they were painted (1943–1953).
An important parameter for quantifying a fractal
pattern’s visual complexity is the fractal dimension,
D. This parameter describes how the patterns
occurring at different magnifications combine to
build the resulting fractal shape D values for various
natural fractal patterns.
Eye-tracks are overlaid on the observed fractal
patterns, which have dimensions of D = 1. 11 (far
left), D = 1.66 (second left), and D = 1.89 (third
left). The final pattern (right) is a colored composite
of four D = 1.6 patterns.
Is Jackson Pollock an artistic enigma? According to our
results, the low D patterns painted in his earlier years
should have more “visual appeal” than the
higher D patterns in his later classic poured paintings.
What was motivating Pollock to paint
high D fractals? Should we conclude that he wanted his
work to be esthetically challenging to the gallery
audience? It is interesting to speculate that Pollock might
have regarded the visually restful experience of a
low D pattern as being too bland for an artwork and
that he wanted to keep the viewer alert by engaging
their eyes in a constant search through the dense
structure of a high D pattern. Speculation over Pollock’s
preference for high D fractals leads us back to the
fundamental question driving this article: why do most
people prefer fractals in the range D = 1.3–1.5?
http://dx.doi.org/10.1177/0301006616633384
http://dx.doi.org/10.1371/journal.pone.0012268
Art images and natural scenes have in common that their radially averaged (1D) Fourier
spectral power falls according to a power-law with increasing spatial frequency (1/f2
characteristics), which implies that the power spectra have scale-invariant properties.
In the present study, we show that other categories of man-made images, cartoons and
graphic novels (comics and mangas), have similar properties.
In conclusion, the man-made stimuli studied, which were presumably produced to evoke
pleasant and/or enjoyable visual perception in human observers, form a subset of all
images and share statistical properties in their Fourier power spectra. Whether these
properties are necessary or sufficient to induce aesthetic perception remains to be
investigated.
20. Biological Sparse Systems
Whywould it matter in visual perception?
http://dx.doi.org/10.1016/j.conb.2004.07.007
http://dx.doi.org/10.1016/j.tins.2015.05.005
http://dx.doi.org/10.1523/JNEUROSCI.0396-16.2016
http://dx.doi.org/10.1098/rsos.160027
We used an algorithm that models the sparseness of the activity
of simple cells in the primary visual cortex (or V1) of humans
when coding images of female faces. Sparseness was found
positively correlated with attractiveness as rated by men and
explained up to 17% of variance in attractiveness. Because V1 is
adapted to process signals from natural scenes, in general, not
faces specifically, our results indicate that attractiveness for
female faces is influenced by a visual bias. Sparseness and
more generally efficient neural coding are ubiquitous, occurring
in various animals and sensory modalities, suggesting that the
influence of efficient coding on mate choice can be
widespread in animals.
The efficient coding strategy is adaptive in at least two ways.
With redundancies discarded, signals are compacted and are
thus more rapidly and precisely processed, which facilitates
memory storing and retrieving [13]. In addition, vision is
remarkably costly: in humans, information coding and
processing within the visual system alone accounts for 2.5–3.5%
of a resting body’s overall energy needs [14]. Because it
requires a limited number of active neurons, sparse coding
therefore allows saving metabolic resources [10,15].
22. Deep learning & Artificial intelligence
Predicting aesthetic value from images
Summary:
How to predict computationally the pleasantness/ popularity/
originality of a given image
• Most of the machine learning work done have focused on “non-
art” images such as Flickr databases which are useful in many
technical systems.
• Very limited work have been done trying to characterize
artistic imagery, and even auto-generate them using generative
models (such as generative adversarial networks, GAN)
• Most of the deep learning work have been devoted of
classifying art movements, recognizing artist and possible
forgeries.
23. 'Classical' Brain analogy of deep learning
http://www.slideshare.net/philipzh/a-tutorial-on-deep-learning-at-icml-2013
Example visualization of an image
classification ConvNet (convolutional n
etworks)
Zeiler and Fergus (2014)
Hierarchical processing (feedforward or recurrent) of information with different layers, lower levels
functioning more as edge detectors, while higher levels have more abstract representations
http://dx.doi.org/10.1126/science.1238406
http://dx.doi.org/10.1016/j.conb.2012.12.008
24. 'Non-art' Aesthetics and Deep learning #1
http://dx.doi.org/10.1145/2647868.2654927
Automated assessment or rating of pictorial aesthetics has many applications. In an image
retrieval system, the ranking algorithm can incorporate aesthetic quality as one of the factors. In
picture editing software, aesthetics can be used in producing appealing polished photographs.
Previous work have formulated the problem as a classification or regression problem where a
given image is mapped to an aesthetic rating, which is normally quantized with discrete values.
Under this framework, the effectiveness of the image representation, or the extracted features,
can often be the accuracy bottleneck
http://dx.doi.org/10.1016/j.image.2016.05.004
In this paper, we classify all images into three categories, namely “scene”, “object” and “texture” ..
Three specific CNNs, namely Scene CNN, Object CNN and Texture CNN, are constructed. The
CNNs learn aesthetic features automatically. Moreover, a single CNN, namely A&C CNN, is also
developed to learn effective features simultaneously for two targets: the aesthetic quality
assessment and the category recognition.
It is shown that the salient region is very important for assessing the aesthetic quality of “object”
images and that the local view is sufficient for assessing “texture” images. In future work, we will
investigate those images in each category that have high aesthetic scores.
http://dx.doi.org/10.1016/j.image.2016.05.009
25. 'Non-art' Aesthetics and Deep learning #2
The Brain-Inspired Deep Networks (BDN) architecture. The input image is first processed by parallel pathways,
each of which learns an attribute along a selected feature dimension independently. Except for the first three simplest
features (hue, saturation, value), all parallel pathways take the form of fully-convolutional networks, supervised by
individual labels; their hidden layer activations are utilized as learned attributes. We then associate those pre-trained
pathways with the high-level synthesis network, and jointly tune the entire network to predict the overall
aesthetics ratings. In addition to the binary rating prediction, we also extend BDN to predicting the rating
distribution, by introducing a Kullback-Leibler (KL)-divergence based loss of the high-level synthesis network.
26. 'Non-art' Aesthetics and Deep learning #3
http://arxiv.org/abs/1605.07699
Many aesthetic models in computer vision suffer from two shortcomings: 1) the low
descriptiveness and interpretability of those hand-crafted aesthetic criteria (i.e.,
nonindicative of region-level aesthetics), and 2) the difficulty of engineering
aesthetic features adaptively and automatically toward different image sets. To
remedy these problems, we develop a deep architecture to learn aesthetically-
relevant visual attributes from Flickr, which are localized by multiple textual
attributes in a weakly-supervised setting.
The pipeline of the proposed CNN-based aesthetic modeling framework (The blue and green arrows denote the training and test phases respectively. The color-
tagged regions indicate visual attributes which are aesthetically pleasing.).
27. 'Non-art' Aesthetics and Deep learning #4
www.cv-foundation.org/openaccess/content_cvpr_2016
Effect on image transformation on photo
aesthetics. (a): Cropping compromises the
composition of the originally well-composed
image that follows rule of thirds. (b): Scaling
distorts the important object. (c): While
padding and scaling keeps the original aspect
ratio, it sometimes leads to the loss of the
image clarity. In this example, the spots on
the ladybug is difficult to see in the padding
result. The added boundaries between the
image and the padding area can also confuse
a deep learning algorithm.
In this paper, we present a composition-preserving
deep ConvNet method that directly learns aesthetics
features from the original input images without any
image transformations. Specifically, our method adds
an adaptive spatial pooling layer upon the regular
convolution and pooling layers to directly handle
input images with original sizes and aspect ratios.
Automatic cropping. We slide a cropping window
through the whole image with the step size of 20
pixels. Each cropping result is scored by our MNA-
CNN method. We show the highest rated cropping
results in (c) and the lowest-rated cropping results in
(d). (b) is a cropping quality map with high values
indicating the locations on the image that our method
suggests a cropping window should be centered at to
create a good cropping result.
Category-aware photo filter recommendation
system. Due to the rapid growth of image filters,
it is difficult for users to choose an ideal filter
efficiently. Meanwhile, we observe that the
selection of image filters is highly related to
image categories (e.g., food, portrait). Hence, we
propose category-aware aesthetic learning by
utilizing our new collected pairwise labeled
image dataset (FACD) for filter recommendation.
http://arxiv.org/abs/1608.05339
28. 'Non-art' Aesthetics and Deep learning #5
http://arxiv.org/abs/1604.04970
http://arxiv.org/abs/1606.01621
Human beings often assess the aesthetic quality of an image coupled with the
identification of the image’s semantic content. This paper addresses the correlation
issue between automatic aesthetic quality assessment and semantic recognition. We
cast the assessment problem as the main task among a multitask deep model, and argue
that semantic recognition task offers the key to address this problem.
Although the proposed multi-task
framework results in state-of-the-
art results on the challenging
dataset, how to perform aesthetic
quality assessment like a human
brain is still an ongoing issue.
Future work is to explore other
possible solutions to efficiently
utilize the aesthetic and semantic
information in a brain-like way.
Another possible trend is to
discover more possible and
potential factors to affect aesthetic
quality assessment.
Real-world applications could benefit from the ability to automatically
generate a fine-grained ranking of photo aesthetics. However, previous
methods for image aesthetics analysis have primarily focused on the coarse,
binary categorization of images into high- or low-aesthetic categories.
In this work, we propose to learn a deep convolutional neural network to
rank photo aesthetics in which the relative ranking of photo aesthetics are
directly modeled in the loss function. Our model incorporates joint learning
of meaningful photographic attributes and image content information which
can help regularize the complicated photo aesthetics rating problem.
29. 'Non-art' Aesthetics and Deep learning #6
“We show that CNN-based approaches outperform the state-of-theart results in
all the 8 tasks. Furthermore, we show that concatenating CNN features learned
from different tasks can enhance the performance in each task. We also show that
concatenating the CNN features learned from all the tasks under experiment does
not perform the best, which is different from what is usually shown in previous
works. Using CNN as a tool to correlate different tasks, we suggest which CNN
features researchers should use in each task.”
[20] F. S. Khan, S. Beigpour, J. V. D. Weijer, and M. Felsberg. Painting-91: a large scale database for
computational painting categorization. Machine Vision and Applications, 25:1385–1397, 2014. Cited by 16
, http://dx.doi.org/10.1007/s00138-014-0621-6
[28] J. Machajdik and A. Hanbury. Affective image classification using features inspired by psychology
and art theory. In Proceedings of the International Conference on Multimedia, pages 83–92, 2010.
Cited by 238, http://dx.doi.org/10.1145/1873951.1873965
http://dx.doi.org/10.1109/WACV.2016.7477616
30. 'Non-art' Aesthetics and Deep learning #7
www.cv-foundation.org/openaccess/content_cvpr_2014
“We propose a number of new and existing computational features, based on aesthetic value and
novelty, for modeling creative micro-videos. We show that groups of features based on scene
content, video novelty, and composition and photographic technique are most correlated with
creative content. We show that specific features measuring order or uniformity correlate with
creative videos, and that creative videos tend to have warmer, brighter colors, and less frenetic, low
volume sounds”
https://arxiv.org/abs/1512.06785
User preference profiling is an important task in modern online social networks (OSN). With the
proliferation of image-centric social platforms, such as Pinterest, visual contents have become one
of the most informative data streams for understanding user preferences. Our approach enables
fine-grained user interest profiling directly from visual contents. For images under the same
label, we reveal intra-categorical variances that traditional classification methods were not able to
capture. We propose a novel distance-metric learning method based on the combination of
traditional-CNN and Siamese Network models
Our experimental results based on
data from 5,790 Pinterest users show
that the proposed method is able to
characterize the intra-categorical
interests of a user with a resolution
that is beyond what a coarse-
grained image classification can do.
Our findings suggest that there is
great potential in finer-grained
user visual preference profiling,
and we hope this paper will fuel
future development of deeper and
finer understanding of users’ latent
preferences and interests.
31. 'Non-art' Aesthetics and Machine learning Fashion #1
http://dx.doi.org/10.1145/2872427.2883037
http://dx.doi.org/10.1007/978-3-319-10590-1_31, Cited by 40
http://www.cv-foundation.org/openaccess/content_cvpr_2016 http://arxiv.org/abs/1608.07444
In this study, computer vision and machine learning techniques were utilized to
made a quantitative study to the influence power of style, color and texture on
the clothing fashion updates. First, three experiments were designed to select
reliable feature descriptors for clothing style, color and texture description,
respectively … Experimental results demonstrated that, on clothing-fashion
updates, the style held a higher influence than the color, and the color held a
higher influence than the texture.
32. 'Non-art' Aesthetics and Machine learning Fashion #2
http://www.cv-foundation.org/openaccess/content_cvpr_2015
In this paper, we analyze the fashion of clothing of a large social website. Our
goal is to learn and predict how fashionable a person looks on a
photograph and suggest subtle improvements the user could make to
improve her/his appeal. We propose a Conditional Random Field model that
jointly reasons about several fashionability factors such as the type of outfit
and garments the user is wearing, the type of the user, the photograph’s
setting (e.g., the scenery behind the user), and the fashionability score.
Importantly, our model is able to give rich feedback back to the user,
conveying which garments or even scenery she/he should change in order to
improve fashionability
Our work is also related to the recent approaches that aim at modeling the
human perception of beauty. In previous work the authors addressed the question
of what makes an image memorable, interesting or popular. This line of
work mines large image datasets in order to correlate visual cues to popularity
scores (defined as e.g., the number of times a Flickr image is viewed), or
“interestingness” scores acquired from physiological studies. In our work, we tackle
the problem of predicting fashionability. We also go a step further from previous
work by also identifying the high-level semantic properties that cause a
particular aesthetics score, which can then be communicated back to the
user to improve her/his look. The closest to our work is Khosla et al. (2013) which
is able to infer whether a face is memorable or not, and modify it such that it
becomes. The approach is however very different from ours, both in the domain
and in formulation. Parallel to our work, Yamaguchi et al. (2014) investigated the
effect of social networks on votes in fashion websites
34. Imaging techniques
Fine-tuning analysis
Published in: Signal Processing Conference, 2011 19th European
Bruno Cornelis ; Ann Dooms ; Jan Cornelis ; Frederik Leen ; Peter Schelkens
http://www.eurasip.org/Proceedings/Eusipco/Eusipco2011/papers/1569424523.pdf
Computer Analysis of Images and Patterns
David G. Stork
Ricoh Innovations and Department of Statistics, Stanford University
http://www.diatrope.com/stork/StorkIP4AI.pdf
35. Multispectral Imaging of art #1
An affordable Multispectral
Imaging System for pigments
mapping on works of art and
archaeology
indiegogo.com
http://chsopensource.org/
Multispectral imaging systems are not invasive and are
successfully used in art examination to map and identify artists’
materials (pigments and binders) and to enhance the reading of
old documents. This is possible thanks to the different
reflectance spectral features that characterize pigments.
Multispectral imaging allows us to examine a painting under different ranges of electromagnetic wavelengths. The
examples above demonstrate the most commonly used imaging methods for examining paintings. Note how each type
of multispectral imaging reveals particular information based on the abilities of that method.
http://www.webexhibits.org/pigments/intro/visible.html
Although a number of companies now offer such
multispectral cameras, Art Innovation (Enschede, The
Netherlands; www.art-innovation.nl) has specifically
targeted this market with its Artist camera.
http://www.vision-systems.com/articles/print/volume-20/issue-3/
http://www.art-innovation.nl/
https://www.hioa.no/eng/employee/rajshr
Color and Imaging Conference, Volume 2015, Number 1, October 2015, pp. 36-40(5)
www.researchgate.net
36. Multispectral Imaging of art #2
The AIC PhD target (left) was accessorized with swatches of UV
and IR fluorescent paints (right) to aid calibration of UVF and IRF
photography. Both targets are necessary for check the correct
shooting and post-processing of all the 8 imaging methods.
Multispectral images
of 56 historical
pigments laid with
gum Arabic on
watercolor paper.
Antonino Cosentino
Heritage Science 2014, 2:8 | DOI: 10.1186/2050-7445-2-8
Miniaturised Fiber Optics Reflectance Spectroscopy (FORS) system (from left
to right): halogen lamp, OceanOptics USB4000 spectrometer and integrating
sphere.
http://dx.doi.org/10.18236/econs2.201410
http://dx.doi.org/10.1016/j.microc.2016.06.020
37. Multispectral Imaging of art #3
1 Conservation Division, National Gallery of Art, 2000-B S Club Drive, Landover, MD 20785
2 The Genomics Institute of the Novartis Research Foundation, 10675 John Jay Hopkins Drive, San Diego, CA 92121
https://www.nap.edu/read/11413/chapter/10
Diffuse reflectance spectra of six blue pigments in powdered form
(dark line) and in oil-bound paint (blue line).
Art & Cultural Heritage
Multispectral Imaging | State-Of-The-Art
Multispectral imaging of artistic and historic works is a
valuable tool to conservators for:
•
Non-invasive characterization
•
Evaluating layers from under-drawings to varnish
•
Revealing watermarks and hidden features
•
Color and pigment analysis
•
Authentication
Multi-Wavelength Analysis | Inside Out
Because different pigments and materials reflect or absorb various
wavelengths (colors of light) differently, a multispectral camera can be
used to reveal and distinguish these. A multispectral camera is able to
investigate these optical properties in a specified spectral range, so the
user can compare findings with known databases of materials, or unveil
hidden features which cannot be seen by the eye. Having this
capability is invaluable in art analysis and restoration projects.
Applications can range from paintings to drawings, pottery, documents,
stamps, textiles and more.
Applications
•
Art conservation & archaeology
•
Cultural heritage & historic works
•
Reveal underdrawings & hidden
features
•
Evaluate surface quality &
restorations
•
Distinguish inks & pigments
•
Document & archive
•
Enhance faded manuscripts &
drawings
•
Detect retouching & repairs
http://pixelteq.com/art-cultural-heritage/
www.artsy.net
38. Lighting setups for art imaging
RAK—Raking light. The painting is illuminated from the side at a very shallow
angle. The craquelure and surface appearance become more distinguishable.
http://colourlex.com/project/multispectral-imaging/
https://youtu.be/LjTprliFuGs
Grazing light
Experts use grazing light to examine paintings in the visible spectrum. Lights are set
up at very shallow angles to the surface of the painting to create what is known as
grazing light or raking light. Grazing light reveals details such as surface defects,
distortions of the support, craquelure, and impasto with great clarity. Grazing light
also increases the depth of heavy, textured paint strokes, such as found in impasto.
This allows art historians to successfully study stroke patterns, making it easier to
observe the manner of the stroke, the direction of the stroke, and the viscosity of
the paint. You can learn a lot about the artist’s technique as well as what he may
have intended to convey to his viewing public by examining his brushstrokes in such
great detail.
Lights can be set up in other ways to divulge even more information about a
painting. For instance, paintings on canvas can be illuminated from behind, which is
known as transmitted light. This can reveal severe paint loss. Transmitted light can be
applied in other situations, such as in the study of signatures, overpaints, crack
patterns in wooden panels, and alterations to both works of art and documentary
artifacts on paper supports.http://www.webexhibits.org/pigments/intro/visible.html
l
If you were to use grazing light to
examine the Raphael fresco above, you
would see the main figures outlined
with deep incisions. Looking carefully at
the shadows, you could easily spot six
areas. Each of these areas is called
“giornata,” which means “a day’s work.”
When painting fresco, the artist added
a thinner, smooth layer of fine plaster
(the intonaco) to the area of wall that he
expected to complete in a day, often
matching the contours of the figures or
the landscape. A layer of plaster
typically required 10 hours to dry; an
artist would begin to paint after one
hour and continue until two hours
before the drying time — providing him
with seven hours of working time.
http://dx.doi.org/10.1111/j.1477-9730.2011.00664.x
39. Art & Deep Learning
The Art and Artificial Intelligence Laboratory at Rutgers:
Advancing AI Technology in the Digital Humanities
The Art and Artificial Intelligence Laboratory at Rutgers is conducting research on the intersection between the two
disciplines. Our aim is to push the envelope of computer vision and artificial intelligence by investigating perceptual and
cognitive tasks related to human creativity. We are focused on developing artificial intelligence and computer vision
algorithms in the domain of art.
https://sites.google.com/site/digihumanlab/home
40. Art work and Deep learning Introduction
http://dx.doi.org/10.1007/978-3-319-46604-0_60
“Understanding the underlying processes of aesthetic perception is
one of the ultimate goals in empirical aesthetics. While deep
learning and convolutional neural networks (CNN) already arrived
in the area of aesthetic rating of art and photographs, only
little attempts have been made to apply CNNs as the underlying
model for aesthetic perception. The information processing
architecture of CNNs shows a strong match with the visual
processing pipeline in the human visual system. Thus, it seems
reasonable to exploit such models to gain better insight into the
universal processes that drives aesthetic perception. This work
shows first results supporting this claim by analyzing already
known common statistical properties of visual art, like sparsity
and self-similarity, with the help of CNNs. We report about
observed differences in the responses of individual layers
between art and non-art images, both in forward and backward
(simulation) processing, that might open new directions of research
in empirical aesthetics.”
Maximizing Art Probability First, we fine-tune a convolutional neural
network to solve the binary classification task artworks vs. non-
artworks. The original DeepDream technique of tries to modify the
image such that the L2-norm of the activations of a certain layer is
maximized. We modify this objective, such that the class
probability for the artworks category is optimized.
Which layers of a CNN show the highest differences
between artwork and all other images? We evaluate
the separation ability of (1) imagenet CNN (2) natural
CNN
and (3) places CNN.
One hypotheses is that a universal model
of aesthetic perception is based on
sparse, i.e. efficient coding, of sensory
input. If activities in the visual cortex can
be coded with sparse representations, they
allow for efficient processing with minimal
energy. Comparing statistics of natural
scenes and visual art showed that these
two categories of images share a common
property related to sparsity in the
representation
The discrimination ability increases for
imagenet CNN and places CNN in later
layers. It is indeed interesting that this is
reflected in the sparsity values as well. Art
images show more sparse
representations at layer fc6 than non-art
images.
Distribution of sparsity scores for art and non-art images computed for the outputs of two layers.
Columns: conv1 vs. conv3, conv1 vs. conv5, conv1 vs. fc6. Smaller values correspond to higher
sparsity.
41. Art work and Deep learning #1
http://arxiv.org/abs/1602.08855
http://dx.doi.org/10.1145/2911996.2912063
To facilitate computer analysis of visual art, in the form of paintings, we introduce Pandora
(Paintings Dataset for Recognizing the Art movement) database, a collection of digitized
paintings labelled with respect to the artistic movement. The database consists of more
than 7700 images from 12 art movements. Each genre is illustrated by a number of
images varying from 250 to nearly 1000. We investigate how local and global features
and classification systems are able to recognize the art movement.
The best achieved performance was by a
combination of pyramidal LBP and Color
Structure Descriptor. One may expect
the addition of GIST to further increase
the performance, but this does not
happen, probably due to the curse of
dimensionality (the features dimension
reaching 800); in such a case a feature
selection method should be used, but we
consider it outside the scope of the
current paper.
Note: Non-deep approach
Digital analysis of art, such as paintings, is a challenging cross-disciplinary
research problem. It has gained much attention recently due to the emergence of
significant amount of visual artistic data on the web.
Computational techniques to manage online large digital art data have several
applications, such as, e.g. art recommendation systems in the tourism industry,
analysis and labeling tools for experts in museums and detection systems to
identify art forgery. In this paper, we investigate the task of automatically
categorizing a painting to its artist and style.
Inspired by the recent success of CNNs, we base our approach on deep features and employ it for
both components: holistic and part-based representations. We use the VGG-16 network pre-trained
on the ImageNet.
The deep network models available at: http://www.robots.ox.ac.uk/~vgg/research/very_deep/
http://dx.doi.org/10.1109/MSP.2015.2406955 | By training a convolutional neural network (PigeoNET)
on a large collection of digitized artworks to perform the task of automatic artist attribution, the
network is encouraged to discover artist-specific visual features.
42. Art work and Deep learning #2
http://dx.doi.org/10.1109/ICIP.2016.7532335
http://dx.doi.org/10.1109/ICIP.2016.7533051
Curators, art historians, and connoisseurs are often
interested in determining the authorship of
paintings. Machine learning and image processing
techniques can assist in this task by providing
noninvasive, automatic, and objective methods. In this
work, we study the automatic identification of Vincent
van Gogh’s paintings using a Convolutional Neural
Network that extracts discriminative visual patterns
of a painter directly from images, and a machine
learning classifier allied with a fusion method in the
final decision process. We divide each painting into
non-overlapping patches, classify them individually,
and then aggregate the outcomes for the final
response. We find out that using the patch with
highest confidence score leads to the best result,
outperforming the traditional voting scheme. We also
contribute with a new and public dataset for van
Gogh painting identification.
For future work, one could experiment with
data augmentation, such as multi-resolution
analysis and density variance, horizontal and
vertical flips, and general affine transformations.
However, such techniques may distort the
discriminative brush stroke characteristics. With
enough data, it would also be interesting to
optimize the weights in a pre-trained CNN and,
ultimately, train a novel architecture from
scratch. Alternatively, another option could
include investigating CNN learning from few
samples The use of Extreme Value Theory
(EVT) in the fusion step also holds promise.
Finally, multi-class and open set approaches to
better handle the particularities between
distinct painters and the relatively small number
of negative samples are also of interest.
Our objectives are two-folds. On one hand, we would like to train an end-to-end
deep convolution model to investigate the capability of the deep model in fine-art
painting classification problem. On the other hand, we argue that classification of
fine-art collections is a more challenging problem in comparison to objects or face
recognition. This is because some of the artworks are non-representational nor
figurative, and might requires imagination to recognize them.
Hence, a question arose is that does a machine have or able to capture
“imagination” in paintings? One way to find out is train a deep model and then
visualize the low-level to high-level features learnt. In the experiment, we
employed the recently publicly available large-scale “Wikiart paintings” dataset that
consists of more than 80,000 paintings and our solution achieved state-ofthe- art
results (68%) in overall performance.
43. Art work and Deep learning #3
cv-foundation.org/openaccess/content_cvpr_201
6
Results show systematic
improvement over state-of-
the-art on transductive single-
and multi-label approaches as
well as other supervised
approaches previously used in
emotion recognition of
abstract paintings. Future
works will focus on extending
the proposed framework to
handle missing features in
order to integrate other
sources of information (e.g.
text) useful for emotional
abstract painting analysis
44. Art work and Deep learning #4
Inspired by the interesting work (Gatys et al. 2015; Cited by 88) that showed the
effectiveness of correlation between feature maps, we transform such correlations into
style vectors, and utilize them to achieve image style classification. We
comprehensively study performance variations brought by correlations in different
layers, performance variations of different correlations, and the idea of inter-layer
correlation. We demonstrated effectiveness of the proposed style vectors through
image style classification and artist classification, as well as performance comparison
with the state of the art. In the future, deeper studies about the essential
characteristics of such descriptors and how to devise better deep features will be
conducted.
Through extensive experiments on image style classification and artist classification, we
demonstrate that the proposed style vectors significantly outperforms CNN
features coming from fully-connected layers, as well as outperforms the
state-of-the-art deep representation.
https://arxiv.org/abs/1508.06576
Leon A. Gatys,1,2,3∗ Alexander S. Ecker,1,2,4,5 Matthias Bethge1,2,4 1Werner Reichardt Centre for
Integrative Neuroscience and Institute of Theoretical Physics, University of Tubingen, Germany ¨
2Bernstein Center for Computational Neuroscience, Tubingen, Germany ¨ 3Graduate School for Neural
Information Processing, Tubingen, Germany ¨ 4Max Planck Institute for Biological Cybernetics, Tubingen,
Germany ¨ 5Department of Neuroscience, Baylor College of Medicine, Houston, TX, USA
Here we introduce an artificial system based on a Deep Neural Network that creates
artistic images of high perceptual quality. The system uses neural representations to
separate and recombine content and style of arbitrary images, providing a neural algorithm for
the creation of artistic images. Moreover, in light of the striking similarities between
performance-optimised artificial neural networks and biological vision, our work offers a path
forward to an algorithmic understanding of how humans create and perceive artistic imagery.
45. Artistic Influence AITheArt and Artificial Intelligence Laboratory at Rutgers: Advancing AI Technology in the Digital Humanities
https://sites.google.com/site/digihumanlab/research/artistic-influence
When we look at a fine-art paining, an expert, or even an average person can infer
information about the style of that paining (e.g. Baroque vs. Impressionism), the
genre of the painting (e.g. a portrait or a landscape), or even can guess the artist who
painted it. People can look at two painting and find similarities between them in
different aspects (composition, color, texture, subject matter, etc.) This is an
impressive ability of human perception for learning and judging complex aesthetic-
related visual concepts, which for long have been thought not to be a logical
process. In contrast, we tackle this problem using a computational methodology, to
show that machines can in fact learn such aesthetic-related concepts.
Frederic Bazille's Studio 9 Rue de la Condamine (left) and Norman
Rockwell's Shuffleton's Barber Shop (right). The composition of both
paintings is divided in a similar way. Yellow circles indicate
similar objects, red lines indicate composition, and the blue square
represents similar structural element. The objects seen -- a fire
stove, three men clustered, chairs, and window are seen in both
paintings along with a similar position in the paintings. After
browsing through many publications and websites, we conclude that
this comparison has not been made by an art historian before.
Measuring similarity between paintings is
fundamental to discover influences, however,
it is not clear how painting similarity might be
used to suggest influences between artists.
The paintings of a given artist can span
extended period of time and can be
influenced by several other contemporary
and prior artists. Therefore, we investigated
several artist distance measures to judge
similarity in their work and suggest
influences. As a result of this distance
measures, we can achieve visualizations of
how artists are similar to each other, which
we denote by a map of artists.
NOTE! t-Distributed Stochastic Neighbor Embedding (t-SNE) might have been a nicer way to
visualize the clusters.
https://lvdmaaten.github.io/tsne/, Cited by 1716
sklearn.manifold.TSNE | Package 'tsne' - CRAN | karpathy/tsnejs: JavaScript t-SNE
Babak Saleh, Kanako Abe, Ravneet Singh Arora, Ahmed Elgammal
https://arxiv.org/abs/1408.3218
46. Artistic Creativity AI #1TheArt and Artificial Intelligence Laboratory at Rutgers: Advancing AI Technology in the Digital Humanities
https://sites.google.com/site/digihumanlab/research/artistic-creativity
For example, the following plot shows the creativity measurement for classical
paintings from Artchive dataset. Horizontal line indicates the time and vertical
axes shows the creativity score computed by our algorithm.
This paper proposes a novel computational
framework for assessing the creativity of creative
products, such as paintings, sculptures, poetry,
etc. We use the most common definition of
creativity, which emphasizes the originality of the
product and its influential value. The proposed
computational framework is based on
constructing a network between creative products
and using this network to infer about the
originality and influence of its nodes. Through a
series of transformations, we construct a
Creativity Implication Network.
This paper will be published in the sixth International
Conference on Computational Creativity (ICCC) June
29-July 2nd 2015, Park City, Utah, USA.
https://arxiv.org/abs/1506.00711
Clearly, it is not possible to judge creativity
based on one specific aspect, e.g. use of
color, perspective, subject matter, etc. For
example it was the use of perspective that
characterized the creativity at certain point of
art history, however it is not the same aspect for
other periods. This highly suggests the need to
measure creativity along different dimensions
separately where each dimension reflects
certain visual aspects that quantify certain
elements of art. The proposed framework can
be used with multiple artistic concepts to
achieve multi-dimensional creativity scoring.
47. Artistic Creativity AI #2
Each point represents a painting. The horizontal axis is the year the painting was
created and the vertical axis is the creativity score (scaled). Only artist names and dates
of the paintings are shown on the graph because of limited space. The red-dotted-
framed painting by Piet Mondrain scored very high because it was wrongly dated in the
dataset to 1910 instead of 1936. See Section 5.2 for a detailed explanation.
Creativity scores for 62K painting from the Wikiart dataset. The horizontal
axis is the year the painting was created and the vertical axis is the scaled
creativity score.
Creativity scores
for 5256 religious
paintings from the
Wikiart dataset
(AD 1410- 1993),
emphasizing
originality in
computing the
creativity sores.
The horizontal
axis is the year the
painting was
created and the
vertical axis is the
scaled creativity
score.
Creativity scores
for 5256 religious
paintings from the
Wikiart dataset
(AD 1410- 1993),
emphasizing
influence in
computing the
creativity sores.
The horizontal
axis is the year the
painting was
created and the
vertical axis is the
scaled creativity
score.
Two dimensional creativity scores for 12310 portrait paintings from the
Wikiart dataset, ranging from 1420 until 2011.
49. Creativity and innovation
From Elgammal and Saleh (2015): There is a historically long and ongoing
debate on how to define creativity. In this section we give a brief
description of some of these definitions that directly relate to the notion we
will use in the proposed computational framework. Therefore, this section is
by no means intended to serve as a comprehensive overview of the subject.
We refer readers to [Taylor, 1988, Paul and Kaufman, 2014] for
comprehensive overviews of the different definitions of creativity.
We can describe a person (e.g. artist, poet), a product (painting, poem), or
the mental process as being creative [Taylor, 1988,
Paul and Kaufman, 2014]. Among the various definitions of creativity it
seems that there is a convergence to two main conditions for a product to
be called “creative”. That product must be novel, compared to prior work,
and also has to be of value or influential [Paul and Kaufman, 2014]. These
criteria resonate with Kant’s definition of artistic genius, which emphasizes
two conditions “originality” and being “exemplary”1
. Psychologists would not
totally agree with this definition since they favor associating creativity with
the mental process that generates the product [Taylor, 1988, Nanay, 2014].
However associating creativity with products makes it possible to argue in
favor of “Computational Creativity”, since otherwise, any computer product
would be an output of an algorithmic process and not a result of a creative
process. Hence, in this paper we stick to quantifying the creativity of
products instead of the mental process that create the product.
Boden suggested a distinction between two notions of creativity:
psychological creativity (P-creativity), which assesses novelty of ideas with
respect to its creator, and historical creativity (H-creativity), which assesses
novelty with respect to the whole human history [Boden, 1990, Cited by 3188
]. It
follows that P-creativity is a necessary but not sufficient condition for H-
creativity, while H-creativity implies P-creativity [Boden, 1990, Nanay,
2014]. This distinction is related to the subjective (related to person) vs.
objective creativity (related to the product) suggested by Jarvie [Jarvie,
1986]. In this paper our definition of creativity is aligned with objective/H-
creativity, since we mainly quantify creativity within a historical context.
1
Among four criteria for artistic genius suggested by Kant, two describe the characteristic of a creative
product “That genius 1) is a talent for producing that for which no determinate rule can be given, not a
predisposition of skill for that which can be learned in accordance with some rule, consequently that
originality must be it’s primary characteristic. 2) that since there can also be original nonsense, its products
must at the same time be models, i.e., exemplary, hence, while not themselves the result of imitation, they
must yet serve others in that way, i.e., as a standard or rule for judging.” [Guyer and Wood, 2000]-p186
http://dx.doi.org/10.1016/j.tree.2015.10.004
In contrast to the siloed research domains of creativity,
innovation, and entrepreneurship, where interdisciplinary
barriers have largely prevented collaboration and
integration, this article focuses on the inextricable, self-
reinforcing linkages between the 3 and the productive
exploitation of those linkages in both education and
practice. Ideally, it is hoped that this discussion will serve
as a catalyst to encourage creativity, innovation, and
entrepreneurship researchers to work together. In so
doing, they can exploit the intersection of these domains,
particularly at the boundaries where the potential to
expand knowledge is rife with opportunity, thereby moving
us closer to integrating and benefiting both theory and
practice.
http://dx.doi.org/10.1037/aca0000015
Published 29 February 2016.DOI: 10.1098/rstb.2015.0182
50. Human creativity #1
http://dx.doi.org/10.1016/j.lindif.2016.09.003
Creative giftedness is represented by a high ability to produce ideas that are
original and valuable in a specific domain or in several domains of work. Moreover,
high levels of creativity are explained by specific processes that are not involved in
high academic achievement. Finally, some researchers have observed cognitive
styles and personality traits that may explain the distinction between high
academic performance and highly creative performance.
http://dx.doi.org/10.1016/j.bandc.2015.09.008
The results revealed that more creative activities were significantly and positively
associated with larger gray matter volume (GMV) in the regional premotor cortex
(PMC), which is a motor planning area involved in the creation and selection of
novel actions and inhibition. In addition, the gray volume of the PMC had a
significant positive relationship with creative achievement and Art scores, which
supports the notion that training and practice may induce changes in brain
structures. These results indicate that everyday creativity is linked to the PMC and
that PMC volume can predict creative achievement, supporting the view that motor
planning may play a crucial role in creative behavior.
The importance of brain connectivity for creativity has been theoretically suggested and
empirically demonstrated. Studies have shown sex differences in creativity measured by
divergent thinking (CMDT) as well as sex differences in the structural correlates of CMDT.
However, the relationships between regional white matter volume (rWMV) and CMDT and
associated sex differences have never been directly investigated. Using rigorous methods,
our findings further supported the importance of brain connectivity for creativity as well as
its female-specific association.
Scholars interested in creative achievement have for years been postulating that
intelligence is a conditio sine qua non for creativity; yet, they tested this hypothesis in a
suboptimal way. This study provides an example of applying the new methodology of
estimating the necessary-but-not-sufficient condition: the NCA. It demonstrates that
intelligence can indeed be perceived as a necessary-but-not-sufficient condition of creative
ability, creative activity, and creative achievement.
http://dx.doi.org/10.1016/j.intell.2016.04.006
51. Human creativity #2
http://dx.doi.org/10.1002/hbm.23246
http://dx.doi.org/10.1038/srep25395
http://dx.doi.org/10.1371/journal.pone.0142567
Moreover, we demonstrate the efficacy of Latent Semantic Analysis as an objective
measure of the originality of ideas, and discuss implications of our findings for the
nature of creativity. Namely, that creativity may not be best described as a stable
individual trait, but as a malleable product of context and perspective.
http://dx.doi.org/10.1016/j.neuroimage.2015.02.002
http://dx.doi.org/10.1038/srep10964
Stimulating creativity has great significance for both individual success and social
improvement. Although increasing creative capacity has been confirmed to be possible and
effective at the behavioral level, few longitudinal studies have examined the extent to which
the brain function and structure underlying creativity are plastic. These results suggest
that the enhancement of creativity may rely not only on the posterior brain regions that are
related to the fundamental cognitive processes of creativity (e.g., semantic processing,
generating novel associations), but also on areas that are involved in top-down cognitive
control, such as the dACC and DLPFC.
Resting-state functional connectivity (RSFC), the temporal correlation of intrinsic activation between
different brain regions, has become one of the most fascinating field in the functional imaging studies. To
better understand the association between RSFC and individual creativity, we used RSFC and the figure
Torrance Tests of Creative Thinking (TTCT-F) to investigate the relationship between creativity measured
by TTCT and RSFC within two different brain networks, default mode network and the cognitive control
network. In conclusion, the current study revealed that the higher creativity, as measured by TTCT-F test,
is related to the decreased RSFC between the MPFC and the precuneus and the increased RSFC between
the left DLPFC and the right DLPFC, which are the nodes belong to the DMN and CCN. These results may
indicate that the altered functional connectivity in the brain is crucial to higher creativity. http://dx.doi.org/10.1093/arclin/acw009
52. Creativity inter-individual differences
http://dx.doi.org/10.1371/journal.pone.0079272
Creativity can be defined the capacity of an individual to produce something
original and useful. An important measurable component of creativity is
divergent thinking. Despite existing studies on creativity-related cerebral
structural basis, no study has used a large sample to investigate the relationship
between individual verbal creativity and regional gray matter volumes (GMVs)
and white matter volumes (WMVs). Modern creativity research is attributed
mainly to Joy Paul Guilford in 1950. Guilford indicated that creative thinking is
the concrete manifestation of individual creativity, and that divergent and
convergent thinking together constitute complete creative thinking, the core of
which is divergent thinking [4]. Divergent thinking refers to the ability of an
individual to develop several solutions to a highly complex open-ended
problem [5].
The relationship between regional GMV and
verbal creativity.
The relationship between regional WMV and
verbal creativity.
Verbal creativity was found to be significantly positively correlated with regional GMV in the left
inferior frontal gyrus (IFG), which is believed to be responsible for language production and
comprehension, new semantic representation, and memory retrieval, and in the right IFG, which
may involve inhibitory control and attention switching. A relationship between verbal creativity and
regional WMV in the left and right IFG was also observed. Overall, a highly verbal creative
individual with superior verbal skills may demonstrate a greater computational efficiency in the
brain areas involved in high-level cognitive processes including language production, semantic
representation and cognitive control.
Studies have shown a mean improvement of creative performance following meditation,
however, differences among individuals have been neglected. We examine whether short-term
integrative body–mind training (IBMT), can improve creative performance and seek to
determine which people are most likely to benefit. Our results support previous findings that
meditation improves creative performance more than relaxation training (RT) does. We
obtained substantial differences between individuals which were correlated with aspects of
their mood and personality. This indicates that differences among people are not due only to
error of measurement but are also predicted by their personality and mood. Taken together, our
study may open up an important avenue for research into the individual differences of the
relationship between meditation and creative performance.
http://dx.doi.org/10.1016/j.lindif.2014.11.019
http://dx.doi.org/10.1371/journal.pone.0146768
The dopaminergic (DA) system may be involved in creativity, however
results of past studies are mixed. We attempted to clarify this putative
relation by considering the mediofrontal and the nigrostriatal DA pathways,
uniquely and in combination, and their contribution to two different
measures of creativity–an abbreviated version of the Torrance Test of
Creative Thinking, assessing divergent thinking, and a real-world creative
achievement index.
Taken altogether, our findings support the idea that human creativity
relies on dopamine, and on the interaction between frontal and striatal
dopaminergic pathways in particular. This interaction may help clarify some
apparent inconsistencies in the prior literature, especially if the genes
and/or creativity measures were analyzed separately.
53. Creative enhancers Psychedelic Substances
http://dx.doi.org/10.1080/02791072.2016.1234090
Developing methods for improving creativity is of broad
interest. Classic psychedelics may enhance creativity; however,
the underlying mechanisms of action are unknown. … Classic
psychedelic use may increase creativity independent of its effects
on mystical experience. Maximizing the likelihood of mystical
experience may need not be a goal of psychedelic interventions
designed to boost creativity.
wired.co.uk
Try explaining this one to your boss.
Cocaine is to Wall Street as LSD is to…Silicon
Valley? Sort of. This week,
Vox published a story from a contributor
rehashing his recent experiences “microdosing”
LSD
time.com/money
http://dx.doi.org/10.1007/s00213-016-4377-8
The present data indicate that ayahuasca enhances
creative divergent thinking. They suggest that ayahuasca
increases psychological flexibility, which may facilitate
psychotherapeutic interventions and support clinical trial
initiatives. The present study has shown that ayahuasca
promotes divergent thinking, an ability which has been
shown to be an important aspect in cognitive therapy
(Forgeard and Elstein 2014). Additional research utilizing a
placebo-controlled experimental design, including
additional creativity measures, is warranted, before results
can be generalized.
https://arxiv.org/abs/1605.07153
http://dx.doi.org/10.1073/pnas.1518377113
There is an actual movement towards increased
health or wellness. People who use it for learning,
improve their learning. One Ivy League student said
he was using microdosing to get through the
hardest math class in the undergraduate curriculum,
and he did wonderfully in the class, and others have
used it for social anxiety.
In the decades that followed its discovery, the magnitude of its effect on science, the arts, and
society was unprecedented. LSD produces profound, sometimes life-changing experiences in
microgram doses, making it a particularly powerful scientific tool.
Here, three complementary neuroimaging techniques: arterial spin labeling (ASL), blood oxygen level-dependent
(BOLD) measures, and magnetoencephalography (MEG), implemented during resting state conditions,
revealed marked changes in brain activity after LSD that correlated strongly with its characteristic psychological
effects. Decreased connectivity between the parahippocampus and retrosplenial cortex (RSC) correlated
strongly with ratings of “ego-dissolution” and “altered meaning,” implying the importance of this particular
circuit for the maintenance of “self” or “ego” and its processing of “meaning.” Strong relationships were also
found between the different imaging metrics, enabling firmer inferences to be made about their functional
significance. This uniquely comprehensive examination of the LSD state represents an important advance in
scientific research with psychedelic drugs at a time of growing interest in their scientific and therapeutic
value
The psychedelic experience is by its very nature highly creative, often
involving generation of a high volume of novel ideas and insights; profuse
visual, auditory, and somaesthetic hallucinations; and intense, widely-
valenced emotional experiences (Dittrich, 1998). These profound
alterations in consciousness have most often been compared with either
psychosis on the negative end of the spectrum (
Vollenweider & Kometer, 2010) or transcendent religious and mystical
experiences on the positive end of the spectrum (Pahnke, 1969)
Artists and other creative individuals have often reported using
psychedelic substances in an effort to enhance creative output or novelty
(Sessa, 2008), and some early experimental work suggested positive
effects along these lines (Harman et al., 1966). There are intriguing
parallels here to dreaming and creative inspiration. As discussed above,
historically speaking, true creativity and inspiration were often seen as a
cooption and overshadowing of the individual self by divine or other
‘higher’ forces (McMahon, 2013). One possibility is that evaluating the
psychedelic experience may be akin to the subsequent evaluation of
creatively-generated ideas (Ellamil et al., 2012), or interpreting and
evaluating one’s dreams the following morning
54. Creative vs cognitive enhancers
In the performance-driven contemporary society,
variety of novel cognitive enhancers have been proposed
A young man I’ll call Alex recently graduated from Harvard. As
a history major, Alex wrote about a dozen papers a semester.
He also ran a student organization, for which he often worked
more than forty hours a week; when he wasn’t on the job, he had
classes. Weeknights were devoted to all the schoolwork that he
couldn’t finish during the day, and weekend nights were spent
drinking with friends and going to dance parties. “Trite as it
sounds,” he told me, it seemed important to “maybe appreciate
my own youth.” Since, in essence, this life was impossible, Alex
began taking Adderall to make it possible.
Alex remains enthusiastic about Adderall, but he also has a
slightly jaundiced critique of it. “It only works as a cognitive
enhancer insofar as you are dedicated to accomplishing the task
at hand,” he said. “The number of times I’ve taken Adderall late
at night and decided that, rather than starting my paper, hey, I’ll
organize my entire music library! I’ve seen people obsessively
cleaning their rooms on it.”
Anjan Chatterjee's, a neurologist at the University of
Pennsylvania, research interests are in subjects like the
neurological basis of spatial understanding, but in the past few
years, as he has heard more about students taking cognitive
enhancers, he has begun writing about the ethical implications
of such behavior. In 2004, he coined the term “cosmetic
neurology” to describe the practice of using drugs developed
for recognized medical conditions to strengthen ordinary
cognition. Chatterjee worries about cosmetic neurology, but he
thinks that it will eventually become as acceptable as
cosmetic surgery has; in fact, with neuroenhancement it’s harder
to argue that it’s frivolous.
http://www.newyorker.com/magazine/2009/04/27/brain-gain
Taylor compares the use of brain stimulation by athletes to eating
carbohydrates ahead of an athletic event, in the hopes of boosting
endurance. “It piggybacks on the ability to learn,” he says. “It's not
introducing something artificial into the body.” But Edwards
worries that the availability of transcranial direct current
stimulation (tDCS) devices will tempt athletes to try “brain
doping”, in part because there is no way to detect its use. “If this
is real,” he says, “then absolutely the Olympics should be
concerned about it.”
http://dx.doi.org/10.1038/nature.2016.19534
wired.com/2014/05http://dx.doi.org/10.1126/science.aad589
3
http://dx.doi.org/10.1038/4501157a
http://dx.doi.org/10.1038/452674a
The most popular of the drugs used by respondents
to Nature 's poll seem to have fairly mild neuroenhancing
effects, says Chatterjee, who calls the massive media interest
in these drugs “neurogossip”. Nevertheless, the numbers
suggest a significant amount of drug-taking among
academics.
https://dx.doi.org/10.3389/fpsyg.2016.00232
55. AI Art The coming era?
"Art should be a slap in the
face. A masterpiece cannot exist
except by struggle"
Rene Magritte (1898-1967)
"An idea that is not dangerous is
unworthy of being called an idea
at all"
Oscar Wilde (1854-1900)
“I have forced myself to
contradict myself in order
to avoid conforming to my
own tastes"
Marcel Duchamp (1887-1968) “I am interested in the
creativity of the criminal
attitude because I recognize in
it the existence of a special
condition of crazy creativity. A
creativity without morals fired
only by the energy of freedom and
the rejection of all codes and
laws. For freedom rejects the
dictated roles of the law and of
the imposed order and for this
reason is isolated.”
Joseph Beuys (1921-1986)Fernand Leger, Mechanical Elements,
1920; oil on canvas, 36 1/8 x 23 1/2 in.
Image courtesy Metropolitan Museum of Art
"Never, even as a child,
would I bend to a rule"
Claude Monet (1840-1926)
56. AI ART from Emulation to art for AI?
Singularity http://www.singularity.com/charts/page70.html
Jürgen Schmidhuber, Point Omega, https://youtu.be/KQ35zNlyG-o
http://dx.doi.org/10.1080/14626268.2016.1147469
Humans will resist to the idea that future art work will be created by artificial intelligence
ultimately without any 'great master' involved.
→ After the transition of art from AI to humans, AI will evolve in making art
for other AI systems humans being incapable of understanding it.
http://dx.doi.org/10.1016/S0004-3702(98)00055-1
techinsider.io, Magenta group introduced at Moogfest
57. Generative artwork with artificial intelligence
Artist and researcher Terence Broad is working on his master's at Goldsmith's computing
department; his dissertation involved training neural networks to "autoencode" movies
they've been fed.
http://boingboing.net/2016/06/02/deep-learning-ai-autoencodes.html
http://www.vox.com/2016/6/1/11787262/blade-runner-neural-network-encoding
https://openai.com/blog/generative-models/
https://arxiv.org/abs/1511.06434
https://github.com/Newmu/dcgan_code
https://github.com/carpedm20/DCGAN-tensorflow