Gespeichert in:
Bibliographische Detailangaben
Hauptverfasser: Kang, Bin, Chen, Bin, Wang, Junjie, Li, Yulin, Zhao, Junzhi, Tian, Zhuotao
Format: Preprint
Veröffentlicht: 2025
Schlagworte:
Online-Zugang:https://arxiv.org/abs/2510.05586
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
_version_ 1866908579200499712
author Kang, Bin
Chen, Bin
Wang, Junjie
Li, Yulin
Zhao, Junzhi
Tian, Zhuotao
author_facet Kang, Bin
Chen, Bin
Wang, Junjie
Li, Yulin
Zhao, Junzhi
Tian, Zhuotao
contents Existing Visual Language Models (VLMs) suffer structural limitations where a few low contribution tokens may excessively capture global semantics, dominating the information aggregation process and suppressing the discriminative features in text-driven image retrieval tasks. To address this, we introduce \textbf{CalibCLIP}, a training-free method designed to calibrate the suppressive effect of dominant tokens. Specifically, in the visual space, we propose the Contrastive Visual Enhancer (CVE), which decouples visual features into target and low information regions. Subsequently, it identifies dominant tokens and dynamically suppresses their representations.In the textual space, we introduce the Discriminative Concept Calibrator (DCC), which aims to differentiate between general and discriminative concepts within the text query. By mitigating the challenges posed by generic concepts and improving the representations of discriminative concepts, DCC strengthens the differentiation among similar samples. Finally, extensive experiments demonstrate consistent improvements across seven benchmarks spanning three image retrieval tasks, underscoring the effectiveness of CalibCLIP. Code is available at: https://github.com/kangbin98/CalibCLIP
format Preprint
id arxiv_https___arxiv_org_abs_2510_05586
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle CalibCLIP: Contextual Calibration of Dominant Semantics for Text-Driven Image Retrieval
Kang, Bin
Chen, Bin
Wang, Junjie
Li, Yulin
Zhao, Junzhi
Tian, Zhuotao
Computer Vision and Pattern Recognition
Existing Visual Language Models (VLMs) suffer structural limitations where a few low contribution tokens may excessively capture global semantics, dominating the information aggregation process and suppressing the discriminative features in text-driven image retrieval tasks. To address this, we introduce \textbf{CalibCLIP}, a training-free method designed to calibrate the suppressive effect of dominant tokens. Specifically, in the visual space, we propose the Contrastive Visual Enhancer (CVE), which decouples visual features into target and low information regions. Subsequently, it identifies dominant tokens and dynamically suppresses their representations.In the textual space, we introduce the Discriminative Concept Calibrator (DCC), which aims to differentiate between general and discriminative concepts within the text query. By mitigating the challenges posed by generic concepts and improving the representations of discriminative concepts, DCC strengthens the differentiation among similar samples. Finally, extensive experiments demonstrate consistent improvements across seven benchmarks spanning three image retrieval tasks, underscoring the effectiveness of CalibCLIP. Code is available at: https://github.com/kangbin98/CalibCLIP
title CalibCLIP: Contextual Calibration of Dominant Semantics for Text-Driven Image Retrieval
topic Computer Vision and Pattern Recognition
url https://arxiv.org/abs/2510.05586