RegionBLIP: A Unified Multi-modal Pre-training Framework for Holistic and Regional Comprehension

1 code implementation • 3 Aug 2023 • Qiang Zhou, Chaohui Yu, Shaofeng Zhang, Sitong Wu, Zhibing Wang, Fan Wang

To this end, we propose to extract features corresponding to regional objects as soft prompts for LLM, which provides a straightforward and scalable approach and eliminates the need for LLM fine-tuning.

Image Comprehension