Datasets
Two dance datasets were released. These datasets are exemplary because they complement quantitative data with the subjective reflections of the participating dancers or choreographers.
Development
“Granular Dance” demonstrates how ML can be combined with a sequence blending method that is inspired by computer music. This method extends the creative possibilities of motion synthesis beyond state of the art (SoA).
“Puppeteering AI” introduces two novel methods for a human dancer to control an artificial dancer in real-time. These methods go beyond SoA since they provide an interaction that is more intuitive than a direct control of a ML model.
RAMFEM is the first ML system that translates motion data into raw audio. This approach goes beyond SoA since it allows to automatically realise motion sonification systems that reflect the idiosyncratic approaches individual dancers.
Expressive Aliens combines reinforcement learning, expressive movement descriptors, and a physics simulation. This approach goes beyond SoA since it can be used to create expressive movements for artificial characters with arbitrary morphologies.
Creation
The creative productions go beyond SoA because of the integration of skills and aesthetic interests of professional performers into the development of simulation-based generative instruments (Strings P) and a ML-based artificial dancer (Artificial Intimacy) and because of the combination of live motion capture, idiosyncratic choreographic principles, and simulation-based behaviours (Embodied Machine).
Impacts
Art
E2-Create makes its main impact on the artistic fields of Dance and Technology, Creative Coding, and Generative Art.
Practitioners in Dance and Technology employed software and sensors to translate their expressivity into music and light and to choreograph and rehearse with artificial dancers.
Creative Coders employed ML models to develop generative musical instruments and interactive systems that detect or generate dance movements.
Practitioners in Generative Art were provided with generative systems that illustrate how bodily creativity can be abstracted and how ML techniques and traditional generative methods can be combined.
Science
E2-Create makes its main impact on the academic fields of Movement and Computing, Human Computer Interaction, and Computational Creativity.
Scientists in Movement and Computing were provided with motion capture recordings of professional dancers, with procedures for deriving higher level movement qualities, and with generative methods for simulating these qualities.
Scientists in Human Computer Interaction were provided with methods for establishing intuitive forms of interaction with ML models and with sensors for exploiting minute body movements as interaction modality.
Scientists in Computational Creativity were provided with an ML model that paves the way for future research on how ML benefits from the creativity employed by dancers.
Public
Through performances and process documentations, the audience learns how digital technology is adopted and developed in dance productions, how generative methods unite technical and artistic ideas.
Through workshop showings, the audience encounters how sensors, sonification, and ML foster creative experimentation in dance, and how dance provides a context and inspiration for artists who work with ML and generative methods.
Through public panel discussions, the audience is informed about the potentials and challenges of artists employing AI for realising works, of teaching AI to artists, and of artists contributing to the development of AI.