WRF4CIR: Weight-Regularized Fine-Tuning Network for Composed Image Retrieval

TL;DR

This paper introduces WRF4CIR, a weight-regularized fine-tuning method for composed image retrieval that mitigates overfitting and improves generalization on limited data.

Contribution

The paper systematically studies overfitting in vision-language models for CIR and proposes a novel weight-perturbation regularization technique during fine-tuning.

Findings

WRF4CIR reduces overfitting in CIR tasks.

The method improves generalization gap across datasets.

Significant performance gains over existing methods.

Abstract

Composed Image Retrieval (CIR) task aims to retrieve target images based on reference images and modification texts. Current CIR methods primarily rely on fine-tuning vision-language pre-trained models. However, we find that these approaches commonly suffer from severe overfitting, posing challenges for CIR with limited triplet data. To better understand this issue, we present a systematic study of overfitting in VLP-based CIR, revealing a significant and previously overlooked generalization gap across different models and datasets. Motivated by these findings, we introduce WRF4CIR, a Weight-Regularized Fine-tuning network for CIR. Specifically, during the fine-tuning process, we apply adversarial perturbations to the model weights for regularization, where these perturbations are generated in the opposite direction of gradient descent. Intuitively, WRF4CIR increases the difficulty of fitting the training data, which helps mitigate overfitting in CIR under limited triplet supervision. Extensive experiments on benchmark datasets demonstrate that WRF4CIR significantly narrows the generalization gap and achieves substantial improvements over existing methods.