CLIP-based knowledge projector for image–text matching

Xinfeng Dong

doi:10.1016/j.ipm.2025.104357

CLIP-based knowledge projector for image–text matching

2025-08-23

0

PRE

AI

DOI：10.1016/j.ipm.2025.104357

Original

Original request for help

Save

Abstract

En 中文

• We put forward a knowledge projector network which regards prior knowledge in CLIP (Radford et al., 2021) as a teacher to guide slot attention generation process. • An adaptive weighted fusion module is used to incorporate global features into slot representations. • An effective similarity calculation method is proposed to compare with fine-grained image–text matching methods. The results indicate that our method outperforms CLIP and the most recent image–text alignment algorithms.

Keywords:

Image–text matching

Multimedia analysis

Slot attention

CLIP-based knowledge projector for image–text matching

Abstract

Journal

Organization