Experimenting with One Million Album Covers

Rising to the challenge to create an image search engine using a corpus of one million album covers,  Professor Trenary of Western Michigan University lead a class project that found many exact matches (same file) and many near matches.

Their algorithm matched some that were not the same because it used rough shape matching, and many images were just of the CD or LP label which matched.

Screen Shot 2015-06-30 at 6.49.39 PM

While not at a point of being ready for production use for the Archive, they wrote a nice report on their findings that might be useful to others.   The Internet Archive hopes to enable many more studies using the data in the collection.

Thank you to Brandon Arrendondo,  James Jenkins, Austin Jones, and Professor Trenary.