Exploiting Category Names for Few-Shot Classification with Vision-Language Models

TL;DR

This paper enhances few-shot image classification by leveraging category names to initialize the classification head in vision-language models, achieving state-of-the-art results on multiple benchmarks.

Contribution

It introduces a novel category name initialization method that significantly improves few-shot classification performance in vision-language models.

Findings

Achieves 87.37% accuracy on ImageNet five-shot classification.

Achieves 96.08% accuracy on Stanford Cars five-shot classification.

Outperforms previous methods on several few-shot benchmarks.

Abstract

Vision-language foundation models pretrained on large-scale data provide a powerful tool for many visual understanding tasks. Notably, many vision-language models build two encoders (visual and textual) that can map two modalities into the same embedding space. As a result, the learned representations achieve good zero-shot performance on tasks like image classification. However, when there are only a few examples per category, the potential of large vision-language models is often underperformed, mainly due to the gap between a large number of parameters and a relatively small amount of training data. This paper shows that we can significantly improve the performance of few-shot classification by using the category names to initialize the classification head. With the proposed category name initialization method, our model obtains the state-of-the-art performance on a number of few-shot image classification benchmarks (e.g., 87.37% on ImageNet and 96.08% on Stanford Cars, both using five-shot learning).