Yanbin Long

Research on Cross-modal Image Retrieval and Image-Text Matching Based on Visual-Language Pre-trained Models

This paper systematically reviews the development of visual-language pre-trained models in the field of cross-modal retrieval, classifies and analyzes existing methods from three dimensions, and deeply analyze three key challenges: fine-grained alignment, noise robustness, and inference efficiency.