Search papers, labs, and topics across Lattice.
The paper investigates overfitting in Vision-Language Pre-trained (VLP) models fine-tuned for Composed Image Retrieval (CIR) and identifies a significant generalization gap. To address this, they propose WRF4CIR, a weight-regularized fine-tuning approach that applies adversarial perturbations to model weights during fine-tuning. Experiments on benchmark datasets show WRF4CIR reduces overfitting and improves retrieval performance compared to existing methods.
Adversarially perturbing model weights during fine-tuning dramatically improves generalization in composed image retrieval, a counterintuitive approach that yields state-of-the-art results.
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.
