LoGoPrompt: Synthetic Text Images Can Be Good Visual Prompts for Vision-Language Models

TL;DR

LoGoPrompt introduces synthetic text images as effective visual prompts for vision-language models, significantly improving performance across various tasks without additional trainable parameters.

Contribution

The paper proposes LoGoPrompt, a novel approach that uses synthetic text images as visual prompts, enhancing generalization and performance in vision-language tasks.

Findings

Outperforms state-of-the-art methods in few-shot learning

Enhances base-to-new generalization

Improves domain generalization across 16 datasets

Abstract

Prompt engineering is a powerful tool used to enhance the performance of pre-trained models on downstream tasks. For example, providing the prompt "Let's think step by step" improved GPT-3's reasoning accuracy to 63% on MutiArith while prompting "a photo of" filled with a class name enables CLIP to achieve $80$\% zero-shot accuracy on ImageNet. While previous research has explored prompt learning for the visual modality, analyzing what constitutes a good visual prompt specifically for image recognition is limited. In addition, existing visual prompt tuning methods' generalization ability is worse than text-only prompting tuning. This paper explores our key insight: synthetic text images are good visual prompts for vision-language models! To achieve that, we propose our LoGoPrompt, which reformulates the classification objective to the visual prompt selection and addresses the chicken-and-egg challenge of first adding synthetic text images as class-wise visual prompts or predicting the class first. Without any trainable visual prompt parameters, experimental results on 16 datasets demonstrate that our method consistently outperforms state-of-the-art methods in few-shot learning, base-to-new generalization, and domain generalization.