CLAP: Isolating Content from Style through Contrastive Learning with Augmented Prompts

TL;DR

This paper introduces a contrastive learning approach with data augmentation to disentangle content from style in vision-language models, improving their generalization, robustness, and performance in zero-shot and few-shot tasks.

Contribution

It proposes a novel method to extract pure content features by integrating image and text augmentation into pre-trained CLIP-like models, enhancing their representation quality.

Findings

Significant improvements in zero-shot classification accuracy.

Enhanced robustness to data perturbations.

Better disentanglement of content and style features.

Abstract

Contrastive vision-language models, such as CLIP, have garnered considerable attention for various downstream tasks, mainly due to the remarkable ability of the learned features for generalization. However, the features they learned often blend content and style information, which somewhat limits their generalization capabilities under distribution shifts. To address this limitation, we adopt a causal generative perspective for multimodal data and propose contrastive learning with data augmentation to disentangle content features from the original representations. To achieve this, we begin with exploring image augmentation techniques and develop a method to seamlessly integrate them into pre-trained CLIP-like models to extract pure content features. Taking a step further, recognizing the inherent semantic richness and logical structure of text data, we explore the use of text augmentation to isolate latent content from style features. This enables CLIP-like model's encoders to concentrate on latent content information, refining the learned representations by pre-trained CLIP-like models. Our extensive experiments across diverse datasets demonstrate significant improvements in zero-shot and few-shot classification tasks, alongside enhanced robustness to various perturbations. These results underscore the effectiveness of our proposed methods in refining vision-language representations and advancing the state-of-the-art in multimodal learning.