-
Notifications
You must be signed in to change notification settings - Fork 125
Open
Description
hello,
I use qdrant to store facenet 512D Embeddings.
Consider person A has two images they have >0.7 Cosine Simalirty while comparing the Embeddings.
When I try to visualize them they appear far apart and they are closer to other embeddings of faces of different people.
But when I use find similar Option they appear closer to each other. ie if i find similar for person A all the other entries for Person A Appear in the search.
ive tried changing the algorithms etc. But still no luck.
How do I fix it or is it an issue of the embeddings. But the embeddings are searched properly via the search method. ie similar ones are only picked.
Metadata
Metadata
Assignees
Labels
No labels
