home  \\  work  \\  contact


<< back



Eigenphoto Extractor




The first three eigenimages of the photos that I took during the Beyond Digital residency in Casablanca.



I was curious whether I could learn something about my photography habits by analyzing the dominant compositional motifs in my photos, so I wrote a script that takes a set of input images and uses principal component analysis to extract "eigenimages" of the sort used in early face recognition algorithms.

The
source code requires NumPy, scikit-learn, and numm.



Some of the input photos for the eigenphotos shown above.




A B&W+R eigenphoto as featured in ARC Magazine.




contact@cosmc.net  ||  @cosmc