I'm looking for a desktop search engine for finding images on my local hard drive with high visual similarity. I've only found paid for ones so far. It could be Linux or Windows software; either would be helpful. A Google Desktop plugin would be nice!
My first use case is being able to organise all the processed versions of a photo based on it's original, or finding the original of a processed image.
Answer
ImgSeek is an open-source project that claims to do this.
DigiKam is apparently trying to add similar functionality.
Pixcavator is not open source, but there is a 30 day free trial.
If you're the DIY coder type, I wound up writing the DB-end of a system that allows phash similarity queries over a fairly large image corpus (10M+ images) in <10 ms/image.
It's part of a larger image deduplication system, and rather tightly coupled, but it's all online here.
Later Edit:
I wound up writing a custom PostgreSQL indexing extension to push fuzzy-image-searching into the database. It's open source here. It scales up to at least 30 million images without significant issue (I have ~32 million images in my index).
No comments:
Post a Comment