Synth$^2$: Boosting Visual-Language Models with Synthetic Captions and Image Embeddings

Open in new window