I love the internet. Only on the internet will you discover a discarded university research project, (which probably got an averagish mark) still there to be discovered. Only on the internet would you be able to peruse the technical difficulties on an optical reader being able to work out what genre an album is just by its cover. Only in such an write-up would the difficulty of a human being able to correctly identify a genre just from cover art be discarded as useful info until the results.
Album Cover-Based Genre Classification – by Matt Hoffman
And yet there seems something heroic about the project, some strand of software that might still last when Skynet takes over the Earth. I like the idea of our robot overlords, as they destroy the last remnants of civilization, being able to recognise that “yes: that is a country album we are destroying”. It is this kind of discernment that will give hope for post-human civilization.
Favourite quote from the experiment:
“That the system does not successfully guess the correct genre a majority of the time should not be too surprising, since it does not have access to the semantic information that human beings do, and even with such information it is completely possible for humans to misclassify an album’s genre if only the cover art is available. Examples of such semantic information include titles and artist names (“25 Country Hits” gives a fairly good indication of the genre of an album, for example) or images of artists (rock musicians tend to look recognizably different from country musicians, but accurately distinguishing between them may require the cognitive ability to analyze fine visual details).”
Also pleased to see the blind faith in the All-Music Guide, and a complete ignorance that the job at hand (optical reading of detailed information) is exactly what bar codes were invented for. Still, a 26% strike rate is not to be sniffed at. Especially when you consider he did not use twee indieist as a genre (two colour photo of age old lost icon – cheers!)