In vbonhomme/eigenfaces: eigenfaces using R

Welcome to this minimal vignette using a minimal version of eigenfaces. More will come.

Images should appear quite ugly but run the code for nicer plots.

Here we go. First of all, be sure to install the very last version of eigenfaces, and also of Momocs.

devtools::install_github("vbonhomme/eigenfaces", build_vignettes=TRUE)
devtools::install_github("vbonhomme/Momocs", build_vignettes=TRUE)

We load both of them

library(Momocs)
library(eigenfaces)
````

We gonna use the dataset 'pain', cheers to [PICS](http://pics.psych.stir.ac.uk/) !
```r
data(pain)
# Let's see what it contains.
pain
class(pain)

eigenfaces uses two classes : afaces and pfaces. They are array of images and PCA objects, respectively.

If you wann see faces rather than text, use plot:

# Im sure you wanted to see faces instead.
plot(pain) # angry face. equivalent to plot(faces, 1)
plot(pain, 81) # happy face

We can extract a data.frame from file names. See lf_structure in Momocs to get a description and a how-to. Here we can cal it to pain directly and it gonna use the dimnames(pain)[[3]] to create the data.frame.

fac <- lf_structure(pain, names=c("ind", "expr"))
head(fac)
````

Here we are. Now, let's calculate a PCA on it. We gonna use the ```PCA``` method from Momocs, that eigenfaces extends. We also pass it with ```fac````.

```r
# if you use Momocs, and you should ;-), code below will look familiar
pain.p <- PCA(pain, fac=fac)

pain.p
head(pain.p$fac)

We can use Momocs plotters

plot(pain.p)
plot(pain.p, "expr") ; axis(1) # we ploti the axis to get an idea of the range (see below)
plot(pain.p, "ind", chull=TRUE, chull.lty=1)

And its boxplot/screeplot plotter:

boxplot(pain.p, "ind")
scree_plot(pain.p)
scree(pain.p)
````

Let's reconstruct faces now from the PCA object:
```r
faces <- PCA2face(pain.p, pos = cbind(seq(-20, 20, 1), 0)) # that's were axis(1) helps

We obtain a afaces object that we can plot.

faces
class(faces)
plot(faces, 10)

We can also explores variation along PC1. Try the following:

for (i in 1:dim(faces)[3]){
  plot(faces, i)
  Sys.sleep(0.1)
}

If you use the package animation, you can even save it as a gif:

library(animation)
saveGIF({
for (i in 1:dim(faces)[3]){
  plot(faces, i)
}
})

Check your working directory for a animation.gif.

More to come.

vbonhomme/eigenfaces documentation built on May 3, 2019, 4:35 p.m.