Unsupervised Vision-and-Language Pre-training Without Parallel Images and Captions

21386

Related Entities