User 984 | 7/3/2015, 7:43:10 PM
I have a stream of images, one arriving every 5 minutes or so, and I would only like to keep the latest image if it is different enough from the previous one. Once I have the images, I'm using them to do classification. I noticed you have some deduplication utilities, but they don't appear to work with images, is that correct?
As a work around, I've gone ahead and trained a deep neural net to perform the classification and then I just extract features from the images and use a distance metric (like gl.distance.euclidean) to determine how close the images are.