One-Shot Learning aims to classify new classes from a single example per class. Methods often leverage metric learning or Siamese networks to compare similarities in embedding spaces. This approach is valuable when data collection is expensive or limited.