Visual-Text Cross Alignment: Refining the Similarity Score in Vision-Language Models

TL;DR

This paper introduces WCA, a method that improves vision-language model alignment by focusing on local image regions and their correspondence with detailed text descriptions, enhancing zero-shot classification accuracy.

Contribution

The paper proposes a localized visual prompting technique and a weighted similarity scoring method to refine cross-modal alignment in vision-language models.

Findings

Significantly improves zero-shot performance across multiple datasets.

Achieves results comparable to few-shot learning methods.

Validates the effectiveness of local visual-text alignment theoretically and empirically.

Abstract

It has recently been discovered that using a pre-trained vision-language model (VLM), e.g., CLIP, to align a whole query image with several finer text descriptions generated by a large language model can significantly enhance zero-shot performance. However, in this paper, we empirically find that the finer descriptions tend to align more effectively with local areas of the query image rather than the whole image, and then we theoretically validate this finding. Thus, we present a method called weighted visual-text cross alignment (WCA). This method begins with a localized visual prompting technique, designed to identify local visual areas within the query image. The local visual areas are then cross-aligned with the finer descriptions by creating a similarity matrix using the pre-trained VLM. To determine how well a query image aligns with each category, we develop a score function based on the weighted similarities in this matrix. Extensive experiments demonstrate that our method significantly improves zero-shot performance across various datasets, achieving results that are even comparable to few-shot learning methods.