The document describes research into developing an automated tool to assist choreographers in their creative process. A survey found that some choreographers are interested in such a tool. A proof-of-concept mobile app was created with different dance styles and rule-based strategies for generating variations. An evaluation found that variations based on an ontology were better than random variations. Presentation through 3D animation was preferred over text, 2D animation or audio. Future work includes developing better dance movement representations and reasoning capabilities.
4. Variation from (Victoria Zhou)
1. Step forward towards corner 2, into
croisé derrière à terre, arms demi-
seconde
2. Plié in 5th change direction to corner 1
3. Relevé derrière, arms in 4th en avant
palms down
4. Passé and change direction to corner 2
place foot on pointe in 5th, arms move
to a low kissing gesture
5. Retiré and back to 5th on pointe with
the front foot, back foot and front foot
arms move to demi-seconde
6. Repeat to the right, and repeat all.
7. Posé coupé effacé towards corner 1,
arms in 4th en avant
8. Posé coupé arms in 4th palms down
9. Step and posé retiré croisé, place back
foot in 5th on pointe, port de bras with
the left arm from 5th en haut to 5th en
avant, end with the arms forward in a
low line, cross the wrists.
19. Cecchetti system
7 elementary movements: [plie (bend), etandre
(stretch), releve (rise), sauter (jump), glisse
(glide), tourne (turn), elancer (dart)]
Positions: 1st, 2nd, 3rd,.. (left right croise)
Facing position (1…8)
Position in space
Direction of movement (de cote, dessous,
dessus, en avant, en arriere, devant, derriere)
Combinations (100+) pas-de-chat, pas-de-
bourre, piroutte
Ballet languages/systems
Based on interview with Marije Koning
20. XML Dance Grammar
Balakrishnan Ramadoss and Kannan Rajkumar. Modeling the Dance Video Semantics using Regular Tree
Automata Fundamenta Informaticae 86 (2008) 175–189 175 IOS Press
23. To what extent can choreographers be supported by
semi-automatic dance analysis and the generation of new
creative elements in choreographies?
24. Method
Questionnaire: How do
choreographers work (with
technologies)
Tool: Proof of concept digital
choreo assistant
Evaluation: Test application to
and different strategies
28. Notations Laban & Benesh
0
5
10
15
20
25
30
Never heard of
it
Cannot work
with it
Other Know Laban Can write both
29. Interest in support in the creative process
Originality, Creativity and Emotion are most important aspects
One very negative sub-group
> Afraid to lose humanity
One positive towards
creative assistance
Two sub-groups:
30. Tool Requirements
based on MoSCoW method
• A dancer must be able to add their dance style to the tool
• A dancer must be able to add their existing choreography to the tool
• The tool must be able to give new suggestions for variations of the
choreography
• The suggestions must be based on different strategies
• The dancer must be able to see the whole choreography at any
moment in time (written)
• The communication of the tool are written dance terms
• The tool must be “easy to use”, which means getting suggestions may
take no longer than 2 minutes
• The tool does have simplified body movements (legs, feet, arms, hands
and head)
31. Proof-of-concept mobile app
3 different dance styles
Ballet (including 78 steps)
Modern dance (including 57 steps)
Street dance (including 31 steps)
Dancepiration – a tool for choreography assistance
32. 4 rule-based strategies for creating variations on
existing choreographies
1. Random step replaced by random other step
2. Random step replaced by ontology-based other step
3. Random steps replaced by multiple strategies
4. Specific step replaced by ontology-based steps
33. Ontology-based variation for the 3 dance styles
.
El Raheb, et al. BalOnSe: Ballet Ontology for Annotating and Searching Video performances.
In Proceedings of the 3rd International Symposium on Movement and Computing (p. 5). ACM, 2016
35. 6 choreography students
Random-based versus Ontology-based
Each dance style is tested 3 times with both strategies per person
Rate original choreography and each variation (10pt scale)
Rate on 5pt Likert scale:
Correctness, Creativity, Helpfulness, Meaningfulness
36. Results
Respondents are positive about the tool
…prefer to choose a specific step to change themselves
… consider creativity in this tool very high (avg 4.2/5)
Correctness is important to improve, it influences other factors the most
50. A significant sub-group of choreographers is interested in and
enthusiastic about automatic choreography support
Needs to be able to understand ‘dance language’
Knowledge representation matters
Style matters
Presentation styles matter -> 3D + dance language
51. Sensing
Representation
+
Reasoning
Presentation
generation
• Motion detection
• Floor sensors
• Move recognition
• Dance movement representation
• Dance choreography representation
• Use of background knowledge
• Pattern detection
• Choreography generation
• Visual presentation
• 3-D animation
• Auditory presentation
Sensing
data
Choreography
variation
Presentation
Choreography
Next level: Representation and Reasoning
Multi-tiered semantic model
Low-level image features
Atomic movements (Labanotation?)
Compound movements (100+ movements)
Emotional content, Socio-cultural layers etc.
Machine Learning for classification and pattern detection
Generative module (automatic choreographer)
Sensing: The environment can be used in different stages of the choreography process. The choreographer can start with a semi-finished dance piece or only have a few dance phrases. The ambient environment detects the current choreography through computer vision, pressure sensors in floor and walls or other modalities. The existing choreography or parts of it are translated into a higher level dance representation language.
Creative Reasoning: The current choreography is transferred to the creative reasoning engine in real time. Here, variations on the choreography are generated. This tool can produce new choreography parts varying spatial positioning, the number of performers, variations of dance phrases, repetitions of dance phrases and differences in expressions of movements. Background knowledge about similar or completely different choreographies as well as different dance styles can also be used to modify the choreography. This background knowledge is also stored in the higher level representation language. All operations on the current choreography can be determined by the user, but a more serendipitous use of the tool is also possible. Multiple variations of one choreography can be generated at the same time.
Presentation: These newly generated choreographies are then presented to the choreographer. For this, 3-D video images of the new choreographies are projected on the walls of the environment. In this manner, the dancer/choreographer can absorb the variations while moving around the environment and he/she can choose to copy entire dance phrases or incorporate only parts of the suggestions. The ambient environment will detect the new movements in real time, creating a creative closed loop where choreographer and the environment work together to create a new dance piece. The tool can also be used offline. The tool then provides the choreographer with an in depth analysis of the current choreography through a dancer-friendly interface.
Al these systems very academic, can be read by a few, mostly used for archival purposes
Already more at normal level of dancer communication
VU + UvA logo + fotootjes
When looking at the results from dance style perspective, it seems that ballet is the worst performing dance style. The correctness of ballet is the lowest in comparison to the other dance styles. It is also the only dance style whereby the random variant performs better than the ontology-based variant.