This repository contains code and data used in the paper Comparative K-Pop Choreography Analysis through Deep-Learning Pose Estimation across A Large Video Corpus:
@article{btdhq2021151,
author = {Peter Broadwell and Timothy R. Tangherlini},
title = {Comparative K-Pop Choreography Analysis through Deep-Learning Pose Estimation across A Large Video Corpus},
journal = {Digital Humanities Quarterly},
volume = {15},
number = {1},
year = 2021
}
The full Python code package documented here is still under development. To try out the pose analysis code in the meantime, we recommend opening the notebook Pose_analysis_examples.ipynb in Google Colab:
Details of the specific functions of each module are available via the project documentation site.
Generally, though, choreo_k
is intended to provide the pose analysis
pipeline routine illustrated in the following figure, in which each
step of the routine can be implemented in one or more different ways,
but each implementation provides roughly equivalent functionality. For
example, different third-party pose detection libraries may be used to
generate pose data, which then can be modified, represented, analyzed
and visualized by any modules that support a specific library’s pose
data output format.