INTENT: Invariance and Discrimination-aware Noise Mitigation for Robust Composed Image Retrieval

2026-04-20

AI Summary

This work addresses the high annotation cost in compositional image retrieval, which induces two types of noise: cross-modal correspondence noise and modality-intrinsic noise. To tackle this challenge, the paper proposes the INTENT framework, which is the first to systematically distinguish and jointly mitigate both noise sources. Specifically, INTENT employs fast Fourier transform (FFT) to implement visual causal intervention, constructing modality-invariant representations that suppress intrinsic noise. Concurrently, it introduces a sample fidelity–based dynamic decision boundary to co-optimize a dual-objective learning scheme, thereby enhancing robustness in cross-modal alignment. Experiments on two mainstream benchmarks demonstrate that INTENT significantly outperforms existing methods, achieving notable advances in both retrieval performance and noise robustness.

Technology Category

• Computer Vision: Multi-modal Vision
• Intelligent Robots: Multimodal Perception & Sensor Fusion
• Machine Learning: Multimodal Learning

Application Category

• Search and Retrieval-Augmented AI: Retrieval-Augmented Generation (RAG) and multi-modal RAG
• User Modeling, Personalization and Recommendation: Fairness-aware retrieval and ranking
• Graph Algorithms and Modeling for the Web: Representation, reconstruction, and subgraph or motif discovery in Web-related graphs

Abstract

Composed Image Retrieval (CIR) is a challenging image retrieval paradigm that enables to retrieve target images based on multimodal queries consisting of reference images and modification texts. Although substantial progress has been made in recent years, existing methods assume that all samples are correctly matched. However, in real-world scenarios, due to high triplet annotation costs, CIR datasets inevitably contain annotation errors, resulting in incorrectly matched triplets. To address this issue, the problem of Noisy Triplet Correspondence (NTC) has attracted growing attention. We argue that noise in CIR can be categorized into two types: cross-modal correspondence noise and modality-inherent noise. The former arises from mismatches across modalities, whereas the latter originates from intra-modal background interference or visual factors irrelevant to the coarse-grained modification annotations. However, modality-inherent noise is often overlooked, and research on cross-modal correspondence noise remains nascent. To tackle above issues, we propose the Invariance and discrimiNaTion-awarE Noise neTwork (INTENT), comprising two components: Visual Invariant Composition and Bi-Objective Discriminative Learning, specifically designed to handle the two-aspect noise. The former applies causal intervention on the visual side via Fast Fourier Transform (FFT) to generate intervened composed features, enforcing visual invariance and enabling the model to ignore modality-inherent noise during composition. The latter adopts collaborative optimization with both positive and negative samples, and constructs a scalable decision boundary that dynamically adjusts decisions based on the loyalty degree, enabling robust correspondence discrimination. Extensive experiments on two widely used benchmark datasets demonstrate the superiority and robustness of INTENT.

Problem

• Composed Image Retrieval
• Noisy Triplet Correspondence
• Cross-modal Correspondence Noise
• Modality-inherent Noise
• Annotation Errors

Innovation

• Composed Image Retrieval
• Noisy Triplet Correspondence
• Visual Invariance
• Discriminative Learning
• Causal Intervention