Learning to Compress Contexts for Efficient Knowledge-based Visual Question Answering

Open in new window