How, by looking at the physical files, to tell the difference between the original photo uploaded to iPhoto, and one of iPhoto's copies of that photo?

Exif is a good approach: it’s possible that iPhoto touches the exif data—I know my raw workflow program adds a tag in exif.

Other than that, if all you want is to eliminate dupes, there are programs (Aperture may) that will do duplicate detection. However, I’m not sure if that does it on image content or just binary.

Are the original created/modified dates still there? Also, file size might help you with some, for example if you reduced the size of the photo in iPhoto.

Kinda looking for a scalable approach… 60,000 files and all. I can’t manually compare one image against another.

If you know of any particular iPhoto-added EXIF tags, that’d be great to have on hand!

Found this nifty utility that can process files based on the EXIF info. Used it to bucket items based on YearTaken_CameraName… if those tokens come up blank or weird, they’re candidates to skip over. But I don’t recall seeing in the instructions that it knows about any iPhoto-specific EXIF tags.

The filesize was a good idea. Salvage operations recovered 146,000 JPGs… and I’ve eliminated 80,000 of those since they were <100KB in size. Got another couple thousand over 10MB in size (one is 780MB!!!) which are also likely to get dropped. So the remaining 60,000 are relatively “good” being in the proper size range and with Year_CameraName info intact.

Will try to stay on the lookout for iPhoto-specific EXIF tags. It appears that iPhoto did preserve the original EXIF tags for the modified photos. Just would like to find the attribute that tells me it’s modified, not original.

Just fyi… digging a little deeper, I downloaded ExifTool. On Macs, it’s a command-line utility for handling EXIF info in JPGs (and other metadata in other filetypes, too).

Here’s what I found out… I found two identical pictures from salvage, one of the them was slightly cropped by iPhoto. Using ExifTool, I got a comprehensive listing of EXIF tags from each file.

The cropped picture had a couple of extra tags… but of main interest to me was the tag named “Host Computer”, which for that particular picture had a value of “Mac OS X 10.3.9”.

So it looks like iPhoto adds a tag identifying the OS version at the time. (You’d think they’d be more interested in tracking iPhoto’s own version, but oh well.)

So I scripted my own utility that runs ExifTool on my 60,000 salvaged pictures, looking only for “Host Computer” tags. On positive hits, it moves the picture file to a different directory, that I can evaluate later.

Been running over nite and it’s nearly done, having identified 16,000 iPhoto-generated copies so far.

Excellent, glad to know you found a solution!