FLEX-CLIP: Feature-Level GEneration Network Enhanced CLIP for X-shot Cross-modal Retrieval

Fuente: arXiv
Gespeichert in:
Bibliographische Detailangaben
Hauptverfasser: Xie, Jingyou, Kuang, Jiayi, Lin, Zhenzhou, Ouyang, Jiarui, Zhao, Zishuo, Shen, Ying
Format: Preprint
Veröffentlicht: 2024
Schlagworte:
Online-Zugang:
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
_version_ 1866929605855674368
author Xie, Jingyou
Kuang, Jiayi
Lin, Zhenzhou
Ouyang, Jiarui
Zhao, Zishuo
Shen, Ying
author_facet Xie, Jingyou
Kuang, Jiayi
Lin, Zhenzhou
Ouyang, Jiarui
Zhao, Zishuo
Shen, Ying
contents Given a query from one modality, few-shot cross-modal retrieval (CMR) retrieves semantically similar instances in another modality with the target domain including classes that are disjoint from the source domain. Compared with classical few-shot CMR methods, vision-language pretraining methods like CLIP have shown great few-shot or zero-shot learning performance. However, they still suffer challenges due to (1) the feature degradation encountered in the target domain and (2) the extreme data imbalance. To tackle these issues, we propose FLEX-CLIP, a novel Feature-level Generation Network Enhanced CLIP. FLEX-CLIP includes two training stages. In multimodal feature generation, we propose a composite multimodal VAE-GAN network to capture real feature distribution patterns and generate pseudo samples based on CLIP features, addressing data imbalance. For common space projection, we develop a gate residual network to fuse CLIP features with projected features, reducing feature degradation in X-shot scenarios. Experimental results on four benchmark datasets show a 7%-15% improvement over state-of-the-art methods, with ablation studies demonstrating enhancement of CLIP features.
format Preprint
id arxiv_https___arxiv_org_abs_2411_17454
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle FLEX-CLIP: Feature-Level GEneration Network Enhanced CLIP for X-shot Cross-modal Retrieval
Xie, Jingyou
Kuang, Jiayi
Lin, Zhenzhou
Ouyang, Jiarui
Zhao, Zishuo
Shen, Ying
Computer Vision and Pattern Recognition
Computation and Language
Given a query from one modality, few-shot cross-modal retrieval (CMR) retrieves semantically similar instances in another modality with the target domain including classes that are disjoint from the source domain. Compared with classical few-shot CMR methods, vision-language pretraining methods like CLIP have shown great few-shot or zero-shot learning performance. However, they still suffer challenges due to (1) the feature degradation encountered in the target domain and (2) the extreme data imbalance. To tackle these issues, we propose FLEX-CLIP, a novel Feature-level Generation Network Enhanced CLIP. FLEX-CLIP includes two training stages. In multimodal feature generation, we propose a composite multimodal VAE-GAN network to capture real feature distribution patterns and generate pseudo samples based on CLIP features, addressing data imbalance. For common space projection, we develop a gate residual network to fuse CLIP features with projected features, reducing feature degradation in X-shot scenarios. Experimental results on four benchmark datasets show a 7%-15% improvement over state-of-the-art methods, with ablation studies demonstrating enhancement of CLIP features.
title FLEX-CLIP: Feature-Level GEneration Network Enhanced CLIP for X-shot Cross-modal Retrieval
topic Computer Vision and Pattern Recognition
Computation and Language
url https://arxiv.org/abs/2411.17454