roaldarbol / tidymocap

🤸 An R package for preparing motion capture data for analysis.

tidymocap

The goal of tidymocap is to tidy data from various motion capture sources. Additionally, augment the poses by calculating kinematics and (some) kinetics.

Installation

You can install the development version of tidymocap like so:

install_github("roaldarbol/tidymocap")

Example

This is a basic example which shows you how to solve a common problem:

library(tidymocap)
data_raw <- tidymocap::anipose_raw
data_tidy <- tidymocap::tidy_anipose(data_raw)
data_augmented <- tidymocap::augment_poses(data_tidy)

tidymocap's Issues

Tidy functions

Use {{ var }} closures throughout.

Make plotting functions

Using plotly for generating stick figures and various data plots.

Generate skeleton in plotly

To make the animation better, make links according to the skeleton file. They're not the same as the joints, so have to be specified differently in the yaml file.

Test with FreeMocap data

Use the data structure from https://github.com/freemocap/freemocap/wiki with their openPoseSkel_3D.

Automatically pull skeletons

Figure out where and how to automatically look for and pull skeletons for the most popular models, e.g. Openpose, MediaPose.or DeepLabCut and automate the process with a GH Action. Otherwise write them yourself.

Make logo

Go back and have a go at the hexSticker package.

Rotational components

Make function that computes rotation around points. Maybe it should take an argument of which joints (a-b-c, DLC style) abour which the rotation is taking place.

How I want the data output:

time, joint, min_likelihood, angle, angle_v, angle_a

Use acos for computing it (https://stackoverflow.com/a/1211243/13240268).

You'll need to pivot_wider() to get all bodypart coordinates on the same line, and then finally pivot_longer() to get 1 observation per joint per line.