I often take a couple of shots of the same thing just to experiment; probably a symptom of growing up shooting digital. I guess you're just banking on randomness preventing dupes from showing too often or manually removing them.
Dupes are definitely possible - though rare in my case. Not only am I randomizing which photos get thumbnailed, I also randomize their order on each page load. I do screen the photos before they get deployed anyways.
Oops, didn't mean it as a callout, just hoped the bug report would be helpful. Probably should have emailed it.
Given the number of browsers you'd need to check, it's completely understandable for a lone developer to miss something like this. Though I do wish people would leave the standard scroll functionality well enough alone.
I learned from a designer friend of mine a long time ago that image resizing to smaller sizes makes them blurry - if you'd just sharpen them a bit it makes a world of difference. It also isn't all that hard to do in Python.
It's fairly slow as-is, taking something like 500ms to locate and render one thumbnail. It might not work the best in a Rails project, but could be used as a pre-processor in a Rake task of some sort.
Not the best - faces might not have the highest entropy in a photo. Existing photo managers (i.e.: iPhoto) already have advanced facial recognition built in, though.
I often take a couple of shots of the same thing just to experiment; probably a symptom of growing up shooting digital. I guess you're just banking on randomness preventing dupes from showing too often or manually removing them.