AI-Generated Choreography - from Solos to Duets

Description

While the fields of technology and dance have historically not often intersected, recent years have seen the advent of AI-generated choreography using models trained on motion capture of a single dancer (https://arxiv.org/abs/1907.05297). This project will expand the state-of-the-art in this intersectional field by exploring duets featuring pairs of dancers, enabling choreography that features authentic interactions between humans & AI models.

Duration

Total project length: 175 hours

Task ideas

Extract pose information from curated videos of dance duets
Train a GNN and/or Transformer model to analyze this data and generate new duet interaction ideas

Expected results

Create a dataset of dynamic point-cloud data corresponding to extracted motion capture poses from videos of dance duets
Train an AI model that can generate the movements of Dancer #2 conditioned on the inputs of Dancer #1 and/or invent new, physically-plausible duet phrases
If time permits: Learn key relationships between parts of the body of each dancer that are integral to the dynamics of the duet
We will collaborate with the original dancers to use the model outputs to inspire new performance material

Requirements

Participants should be comfortable with standard data science software including Python, Git, Numpy, Matplotlib, and Pandas. Previous experience in Machine Learning, either in TensorFlow or PyTorch, is preferred. While previous experience in dance or the performing arts is not needed, an interest in the artistic and open-ended aesthetic dimensions of the project is required. Strong interpersonal & communication skills are essential.

Project difficulty level

Hard

Test

Please use this link to access the test for this project.

Mentors

Mariel Pettee (Lawrence Berkeley National Laboratory)
Ilya Vidrin (Northeastern University)

Please DO NOT contact mentors directly by email. Instead, please email human-ai@cern.ch with Project Title and include your CV and test results. The mentors will then get in touch with you.

Corresponding Project

ChoreoAI