Foundation models with rich knowledge can boost open-world video recognition through a generic knowledge transfer pipeline named PCA.
