I have read the paper and wondering what features (meta-embedding or direct feature) did you use as a face recognition embedding.
Since megaface is an open set, each meta-embedding would be a close-zero vector, so it may not be right to use meta-embedding as the face recognition embedding, am i right?
Hi,
I have read the paper and wondering what features (meta-embedding or direct feature) did you use as a face recognition embedding.
Since megaface is an open set, each meta-embedding would be a close-zero vector, so it may not be right to use meta-embedding as the face recognition embedding, am i right?