Category: Art Open Data

If we save the data of Roger de Piles’ scores for artists to a csv file we can load them into R:

## Load the tab separated values for the table of artist scores
scores<-read.csv("./scores.csv",
colClasses=c("character", "integer", "integer", "integer",
"integer"))
## Replace NA values with zero
nas<-which(is.na(scores), arr.ind=TRUE)
scores[nas[1], nas[2]]<-0
## Create the total score
scores<-cbind(scores, Total=apply(scores[2:5], 1, sum))

This allows us to find the lowest and highest scores:

## Min, max of each score
scoreMinMax<-function(scores, column){
lowest<-min(scores[column])
cat(column, "\nMin (", lowest, "): ", sep="")
cat(scores$Painter[scores[column] == lowest], sep=", ")
highest<-max(scores[column])
cat("\nMax (", highest, "): ", sep="")
cat(scores$Painter[scores[column] == highest], sep=", ")
cat("\n")
}

> scoreMinMax(scores, "Composition")
Composition
Min (0): Guido Reni, Gianfrancesco Penni
Max (18): Guercino, Rubens
>
> scoreMinMax(scores, "Drawing")
Drawing
Min (6): Giovanni Bellini, Lucas van Leyden, Caravaggio, Palma il Vecchio, Rembrandt
Max (18): Raphael
>
> scoreMinMax(scores, "Colour")
Colour
Min (0): Pietro Testa
Max (18): Giorgione, Titian
>
> scoreMinMax(scores, "Expression")
Expression
Min (0): Jacopo Bassano, Giovanni Bellini, Caravaggio, Palma il Vecchio, Gianfrancesco Penni
Max (18): Raphael
>
> scoreMinMax(scores, "Total")
Total
Min (23): Gianfrancesco Penni
Max (65): Raphael, Rubens
>

Cluster the artists:

## Clustering Utilities
clustersNames<-function(clusters, names){
clusterCount<-length(clusters$size)
clusters.works<-lapply(1:clusterCount,
function(cluster){
names[clusters$cluster == cluster]})
}
printClustersNames<-function(clustersNames){
clusterCount<-length(clustersNames)
for(cluster in 1:clusterCount){
cat("Cluster", cluster, ":",
paste(unlist(clustersNames[cluster]), collapse=", "),
"\n\n")
}
}
## Cluster based on the numeric scores. 8 = 2x2x2 (Low/High)
clusters<-kmeans(scores[2:5], 8)
names<-clustersNames(clusters, scores$Painter)
printClustersNames(names)

Cluster 1 : Correggio, Rembrandt, Van Dyck
Cluster 2 : Andrea del Sarto, Federico Barocci, Daniele da Volterra, Guercino, Lucas Jordaens, Giovanni Lanfranco, Otho Venius, Perin del Vaga, Primaticcio, Francesco Salviati, Taddeo Zuccari
Cluster 3 : Charles Le Brun, Il Domenichino, Giulio Romano, Leonardo da Vinci, Eustache Le Sueur
Cluster 4 : I Carracci, Raphael, Rubens, Vanius
Cluster 5 : Guido Reni, Gianfrancesco Penni
Cluster 6 : Jacopo Bassano, Giovanni Bellini, Caravaggio, Murillo, Palma il Vecchio
Cluster 7 : Sebastian Bourdon, Cavalier D'Arpino, Albrecht Dürer, Lucas van Leyden, Michelangelo, Il Parmigianino, Pietro Testa, Federico Zuccari
Cluster 8 : Abraham van Diepenbeeck, Giorgione, Giovanni da Udine, Holbein, Jacob Jordaens, Palma il Giovane, Sebastiano del Piombo, Teniers, Tintoretto, Titian, Veronese

And graph the scores:

## Stacked bar chart
## Allow room for names at bottom and legend at right
## 7 is from trial and error
par(xpd=T, mar=par()$mar+c(7,0,0,7))
barplot(t(as.matrix(scores[2:5])), names.arg=scores$Painter,
main="Roger de Piles' Ratings", col=rainbow(4), las=2, border=NA)
## Position legend in right margin
## 60 is from trial and error
legend(60, 60, names(scores[2:5]), fill=rainbow(5), cex=0.75)

Dissecting the Canon: Visual Subject Co-Popularity Networks in Art Research

This is a well-defined statistical study of the art historical literature of a particular period. It counts the number of times that ancient artworks are mentioned in Renaissance art literature. By measuring the popularity and co-popularity of artworks it uncovers several interesting facts.

Firstly, canons are identical with the most popular items over a distribution of popularity. Secondly, sub-tails of genres or subjects have broadly the same properties as the main long tail of which they are a sub-tail. And thirdly the co-popularity of otherwise unrelated monuments may be a product of their spatial proximity at the time they were documented in the renaissance.

These facts are interesting in themselves and indicate further possibilities for research. They are also of use to more theoretical or social art historical approaches.

(I originally posted about this here.)