Fetching the paper…

SemVLP: Vision-Language Pre-training by Aligning Semantics at Multiple Levels · Around