# BLIP-2

Connects a frozen image encoder to a frozen language model through a lightweight query transformer.

## What it is used for

Image captioning and visual question answering.