CORL: Compositional Representation Learning for Few-Shot Classification | ComputerVisionFoundation Videos | Podwise