You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
DeepFace currently supports cosine distance, Euclidean distance, and Euclidean L2 distance for face embedding comparisons. To enhance distance metric options, we should add angular distance, which is based on the angle between two embeddings. This metric is particularly useful for spherical embeddings and provides an alternative to cosine distance.
Formula
angular_distance=np.arccos(similarity) /math.pi
Additional Info
No response
The text was updated successfully, but these errors were encountered:
Description
DeepFace currently supports cosine distance, Euclidean distance, and Euclidean L2 distance for face embedding comparisons. To enhance distance metric options, we should add angular distance, which is based on the angle between two embeddings. This metric is particularly useful for spherical embeddings and provides an alternative to cosine distance.
Formula
Additional Info
No response
The text was updated successfully, but these errors were encountered: