Image Matching

I’ve been working on videos the past two weeks, spending a lot of time in a bubble examining clips in a sort of detail that is probably excruciating for everyone else. I like this sort of in depth analysis of what I look at. It also brings up how flawed we are in how we process images online. We generally assign a bunch of words, meaning and descriptors of an image instead of trying to mathematically or logically sort images. Google gets it right by being able to match images, but their similar image algorithm would be much more useful if we could upload an image and say “match that!” It’s further complicated when you factor in the 60 images a second that you get with a video. It would be great to be able to upload a clip and get a bunch of clips, YouTube videos that match not only the actions but the content as well. Want to see Devo’s live performance of Whip It? Just whip up a video of you playing Whip It on Rock Band 3, upload it and get matches for the band, but others also playing Devo on Rock Band.

Crap Detector Part 2

Continuing on with the theme yesterday I wanted to add that Howard Rheingold has succinctly written a piece on crap detection, which pretty closely mimics what I’ve said in the last year of my Searching the Internet course. I’ll embed last years’ video lectures from that week’s work.

We work on the same principles, in fact when I teach this face-to-face, I try to accentuate that you have to think like a detective or private investigator; build a case for or against this website’s information. That was interesting because I’ve had cops in my class who said that investigative technique is a lot like taking a bunch of disparate pieces and putting them together is a lot of what detective work consists of. I’m glad we’re on the same path.