Check if EXIF information contains unique ID and use it for duplicate detection about photoprism HOT 9 CLOSED

I would just use a phash. This produces a hash that should withstand scaling and minor amounts of filtering. At best you eliminate duplicates. At worst, you still have them. Either way, you'll eliminate files resulting from simple copies/renaming on disk.

I wrote a Go implementation a couple of years ago that produced identical results to the other implementations at http://blockhash.io . They never updated the website, but it works all the same.

https://github.com/dsoprea/go-perceptualhash

It's not perfect, but it's as close as you're gonna get without some research guys.

from photoprism.

Was also going to look into the sha1 fingerprints we do for all files... my feeling is it adds significant overhead, maybe we can find a faster solution that is also reliable. It is used to determine, if a file needs to be indexed again and if new thumbs are needed.

from photoprism.

That makes more sense. You're not (at this stage) concerned with associating all permutations.

from photoprism.

The light and color maps we have for all photos are some kind of perceptual hash... we want to use that later to easily find similar images.

from photoprism.

That's interesting. Presumably, you'd have to layer them or otherwise marry them together first.

from photoprism.

See https://github.com/photoprism/photoprism/blob/develop/internal/photoprism/colors.go

from photoprism.

We now extract Document and Instance ID via exiftool and use it for grouping related files with different names.

from photoprism.

What if those aren't present (since they're nonstandard)?

from photoprism.

Exif may contain an ImageID, typically a hash or UUID. Not extremely common though. Document and Instance ID can be found in XMP metadata.

from photoprism.