Label Propagation for Zero-shot Classification with Vision-Language Models

TL;DR

This paper introduces ZLaP, a label propagation method leveraging graph structures of unlabeled data with vision-language models, improving zero-shot classification performance especially with unlabeled data.

Contribution

The paper proposes ZLaP, a novel label propagation approach tailored for graphs with text and image features, enhancing zero-shot classification with unlabeled data.

Findings

ZLaP outperforms recent methods on 14 datasets.

The method effectively utilizes graph structure and geodesic distances.

Efficient inductive inference is achieved through dual solutions and sparsification.

Abstract

Vision-Language Models (VLMs) have demonstrated impressive performance on zero-shot classification, i.e. classification when provided merely with a list of class names. In this paper, we tackle the case of zero-shot classification in the presence of unlabeled data. We leverage the graph structure of the unlabeled data and introduce ZLaP, a method based on label propagation (LP) that utilizes geodesic distances for classification. We tailor LP to graphs containing both text and image features and further propose an efficient method for performing inductive inference based on a dual solution and a sparsification step. We perform extensive experiments to evaluate the effectiveness of our method on 14 common datasets and show that ZLaP outperforms the latest related works. Code: https://github.com/vladan-stojnic/ZLaP