Unexpected Similarity results

[Carol C. Kankelborg]

I ran a similarity scan on my entire library of 90K-ish photos and got some weird results.
The settings are: Restrict search “Only selected tabs”; Similarity range: 100%-100%; Limit search “None”

Previous scans, using just a subset of my library performed as expected. It found truly identical images.


I sorted my results by number of items. At the bottom are two reference images with 12 and 32 matches respectively. The alleged similar images are very dissimilar in both content and pixel dimensions. The one with 12 matches is below (assuming I can send an image to this list). There are three all-white JPEGs and the other 9 are nine different B&W scans in XCF format. 

The image of 32 matches include a bunch of what appear to be all black frames of JPG and PNG and NEF. The remaining images are PNG clipart with transparent backgrounds. The reference image is one of the clipart PNGs. 

There are about 100 images with 4 to 8 matches. A quick skim of those indicates they are all correct.

What is going on with the 12 and 32 match images? Is this a known issue? Is this a bug I should report? 

Thanks,
Carol


-------------- next part --------------
A non-text attachment was scrubbed...
Name: Screenshot 2025-05-11 at 18.56.32.png
Type: image/png
Size: 230424 bytes
Desc: not available
URL: <http://mail.kde.org/pipermail/digikam-users/attachments/20250511/c504713e/attachment-0001.png>