Skip to content

Instantly share code, notes, and snippets.

@pranjalg2308

pranjalg2308/cosine_triplet_loss.py

Last active May 25, 2020 14:35
Show Gist options
  • Save pranjalg2308/6bfd49feec0cb5f50ade5fe3d89fec53 to your computer and use it in GitHub Desktop.
Save pranjalg2308/6bfd49feec0cb5f50ade5fe3d89fec53 to your computer and use it in GitHub Desktop.
Download ZIP
Implementation for Cosine similarity for triplet loss in keras
Raw
cosine_triplet_loss.py
# Cosine Similarity
def triplet_loss(alpha, emb_dim):
def loss(y_true, y_pred):
anc, pos, neg = y_pred[:,:emb_size], y_pred[:,emb_size:2*emb_size], y_pred[:,2*emb_size:]
distance1 = tf.keras.losses.cosine_similarity(anc, pos)
distance2 = tf.keras.losses.cosine_similarity(anc, neg)
return tf.keras.backend.clip(distance1 - distance2 + alpha, 0., None)
return loss
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment