CLIP in Mirror: Disentangling text from visual images through reflection

Open in new window