Metric Learning

Angular Loss

Loss function that optimizes the angles between embedding vectors rather than their Euclidean distances, offering better scale invariance and improved class separation. It is particularly useful for facial recognition tasks.

← Indietro