README.md

In tsostarics/ftpals:

First Take Palettes

This package provides color palettes based on Sony’s THE FIRST TAKE video series. This package is not affiliated with Sony Music Entertainment Japan whatsoever.

ftpalsというパッケージのことはSMEやTHE FIRST TAKEに提携していません。

Current counts:

| type | n | |:---------------|----:| | THE FIRST TAKE | 135 | | THE HOME TAKE | 16 |

Goals

I made this as a side project to learn more about working with colors in R. Specifically I wanted to do the following:

• Learn more about working with image data in R (but.. not using magick)
• Try to use clustering as a means to make color palettes
• Try to make those color palettes color blind friendly procedurally

There are packages that provide functionality to create color palettes from images already, such as colorfindr and paletter but I wanted to try something new myself.

I reimplemented the genetic algortihm implementation in this paper:

Troiano, Luigi & Birtolo, Cosimo & Miranda, Maria. (2008). Adapting palettes to color vision deficiencies by genetic algorithm. GECCO’08: Proceedings of the 10th Annual Conference on Genetic and Evolutionary Computation 2008. 1065-1072. 10.1145/1389095.1389291.

As a result I learned about genetic algorithms using the GA package:

Luca Scrucca (2013). GA: A Package for Genetic Algorithms in R. Journal of Statistical Software, 53(4), 1-37.

Luca Scrucca (2017). On some extensions to GA package: hybrid optimisation, parallelisation and islands evolution. The R Journal, 9/1, 187-206.

I used thumbnails from THE FIRST TAKE’s YouTube uploads as a test case, as they often have a nice variety of colors from participants’ clothes and the characteristic bar the studio overlays onto each image. That said, only videos from THE FIRST TAKE or THE HOME TAKE series were used, not THE FIRST TAKE FES.

Documentation

The source files used for scraping the video metadata are available in the R/scrape_utils directory and prepended with scrape_. The files used to implement the genetic algorithm are similarly in R/GA_utils and prepended with GA_. Note that these aren’t exported by the package, but they’re available to look at.

I won’t strongly commit to the actual effectiveness of every palette, but I think some turned out nice!

Installation

You can install the released version of ftpals from github with:

devtools::install_github("tsostarics/ftpals")

Palette Examples

All palettes are provided by the ftpals::first_takes table. This provides additional metadata such as videoId, artist name, and song title. Note that some artist/song names are given in Japanese while some are given in Romaji, this is just how the data got scraped. It seems like if an English translation is provided for a video, then download requests from the United States will pull the translation instead of the original Japanese.

show_pals(i) will show the two palettes available for a given index in the first_takes table. The original sampling of colors, first_takes$palette, is shown on top, and the “optimized” palette, first_takes$optimized_palette, is shown on the bottom for comparison. The “optimized” palette should be a little better for those with colorblindness, but effectiveness varies by palette. Due to the implementation the optimized palette tends to have these qualities:

• The first color is less white than the original sample
• The second color is less black than the original sample
• The palette is reordered by alternating luminance, so the hue ordering sometimes changes
• The palette is often somewhat less saturated than the original palette

Here is an example from this video. You can use the plot_ftpalette() function to plot the palettes with the thumbnail the colors were sampled from. Or, you can use show_pals to see the two palettes with their hex codes.

library(ftpals)
library(ggplot2)
plot_ftpalette(91, show_thumbnail = T, use_theme = ggplot2::theme_minimal)

show_pals(91) # 病身を噛む by ずっと真夜中でいいのに

Here are a few other examples.

show_pals(27) # ミカヅキ by さユり

show_pals(42) # sabotage by 緑黄色社会

show_pals(23) # BL by QUEEN-BEE

show_pals(take = 86) # オーケストラ by AiNA THE END

Here’s an example with a plot. You can pick the palette you want by directly indexing the palette or optimized palette you want. Notice there’s also a bar_color column you can access for an additional highlight color (it tends to be a very saturated color).

diamonds %>% 
  ggplot(aes(x = cut, y = carat, fill = cut)) +
  stat_summary(fun = 'mean',
               geom = 'bar',
               color = 'black') +
  scale_fill_manual(values = first_takes$optimized_palette[[23]]) +
  geom_hline(yintercept = .75,
             size = 2, 
             linetype = "dashed", 
             color = first_takes$bar_color[[23]]) +
  theme_minimal()

Scale_ functions

You can use scale_fill_ftake() and scale_color_ftake() to set colors, too. They take either an index, or a type character denoting which series of videos to use and a take for which specific video to use. If you want to use palettes from THE HOME TAKE, you should opt for the type/take method. If you want to use videos from THE FIRST TAKE, index works fine but there is no video number 70 (either it was accidentally skipped or later taken down), so the index-take number correspondence is off by one after 69.

diamonds %>% 
  ggplot(aes(x = cut, y = carat, fill = cut)) +
  stat_summary(fun = 'mean',
               geom = 'bar',
               color = 'black') +
  scale_fill_ftake(index = 24) +
  geom_hline(yintercept = .75,
             size = 2, 
             linetype = "dashed", 
             color = barcolor(24)) +
  theme_minimal()

diamonds %>% 
  ggplot(aes(x = cut, y = carat, fill = cut)) +
  stat_summary(fun = 'mean',
               geom = 'bar',
               color = 'black') +
  scale_fill_ftake(index = 24, order = 6:1) +
  geom_hline(yintercept = .75,
             size = 2, 
             linetype = "dashed", 
             color = barcolor(24)) +
  theme_minimal()

diamonds %>% 
  ggplot(aes(x = cut, y = carat, fill = cut)) +
  stat_summary(fun = 'mean',
               geom = 'bar',
               color = 'black') +
  scale_fill_ftake(palette = 2, type = "H", take = 10) +
  geom_hline(yintercept = .75,
             size = 2, 
             linetype = "dashed", 
             color = barcolor(type = "H", take = 10)) +
  theme_minimal()

tsostarics/ftpals documentation built on Dec. 23, 2021, 12:59 p.m.