README.md

In javierluraschi/hexkart: kartsim: Kart simulation for model training

kartsim: Kart simulation for model training

kartsim provides a kart simulation to render and capture kart training data using 'HTML' widgets or 'Shiny' applications. It can be used to: Collect kart playing styles, create test cases for image classification and render trained models over an interactive kart simulation.

Getting Started

Install kartsim from CRAN then launch the imulation as follows, use arrow keys to turn left or right:

library(kartsim)
kartsim_play()

To capture a kart session simply run:

kartsim_capture()

This will create a snapshot under capture/ with labels: ###-left.png, ###-forward.png or ###-right.png that you can use for training. Once trained, you can use kartsim_control() to feed predictions back.

We can feed a random direction to this simulation by running:

labels <- c("left", "forward", "right")
kartsim::kartsim_control(function(image, direction) {
  sample(labels, 1)
})

Training

Training this simulation is beyond the scope of this package, but the following github repos can be used for further reading. Feel free to send a PR to this repo if you want your models to be listed here:

• github.com/javierluraschi/kartmodels

javierluraschi/hexkart documentation built on May 6, 2019, 8:33 p.m.