DeepDance: Full-Body Motion Capture of Improvised Dance with Audio (2019)

Nine dancers were recorded individually, improvising to six different audio files, resulting in 162 motion capture files. Full-body motion was captured at 250Hz using a Qualisys infra-red optical system at the University of Oslo in October 2019. The dataset was collected by Benedikte Wallace et al. for use as training data in deep learning for motion generation.

Use Cases

• Training generative models for dance motion based on audio stimuli described in the dataset.
• Analyzing patterns in human improvisation and movement based on standardized body segment data.
• Developing cross-modal models that synchronize motion capture data with corresponding audio files.
• Benchmarking motion prediction algorithms using the high-frequency (250Hz) capture data.

Strengths

• Includes 162 motion capture files from nine dancers, providing multiple performances per musical piece.
• Motion data was captured at a high frequency of 250Hz, offering detailed temporal resolution.
• Data is post-processed and standardized, with body segments normalized and root markers set at the origin.
• Multimodal dataset includes corresponding audio .wav files for each motion capture session.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count and dataset size are unknown, which may limit suitability assessment.
• Last update date is unknown; freshness unverified.

Provenance

Source

University of Oslo

Collection Method

Recorded using a Qualisys infra-red optical motion capture system.

Time Range

October 2019

Geography

Oslo, Norway (University of Oslo)