Composed Image Retrieval
Overview
Composed image retrieval studies how to find a target image from a query that combines a reference image with an instruction, such as changing an outfit while preserving its style. Methods learn joint image–text embeddings from paired examples, or adapt pretrained vision-language models with projection, interpolation, or instruction-based composition; reranking and interactive clarification can further refine results. Work spans fashion and product search as well as medical and other domain-specific retrieval, with both supervised training and zero-shot approaches that use pretrained models without task-specific training. A central challenge is balancing broad semantic understanding with fine-grained attribute and instance fidelity, while ensuring benchmarks test genuine multimodal composition rather than shortcuts and remain robust to ambiguity and noisy labels.
Latest papers
66 papers4 papers in the last four weeks, up 33% on the four weeks before. 0.0% of all new papers.
Mixed-Modality Dual Face-Hair Retrieval
CIRCLED: A Multi-turn CIR Dataset with Consistent Dialogues across Domains
Resolving Ambiguity in Composed Image Retrieval via Calibrated Interaction
Benchmarking Composed Image Retrieval for Applied Earth Observation
DeliCIR: Deliberative Test-Time Evolutionary Hierarchical Multi-Agents for Composed Image Retrieval
STiTch: Semantic Transition and Transportation in Collaboration for Training-Free Zero-Shot Composed Image Retrieval
Do Composed Image Retrieval Benchmarks Require Multimodal Composition?
Decoupling Endpoint and Semantic Transition Learning for Zero-Shot Composed Image Retrieval
TEMA: Anchor the Image, Follow the Text for Multi-Modification Composed Image Retrieval
ConeSep: Cone-based Robust Noise-Unlearning Compositional Network for Composed Image Retrieval
hard noise'' (i.e., the reference and target images are highly similar but the modification text is incorrect), poses a unique challenge to existing Noise Correspondence Learning (NCL) methods because it breaks the traditional small loss hypothesis''. We identify and elucidate three key, yet overlooked, challenges in the NTC task, namely (C1) Modality Suppression, (C2) Negative Anchor Deficiency, and (C3) Unlearning Backlash. To address these challenges, we propose a Cone-based robuSt noisE-unlearning comPositional network (ConeSep). Specifically, we first propose Geometric Fidelity Quantization, theoretically establishing and practically estimating a noise boundary to precisely locate noisy correspondence. Next, we introduce Negative Boundary Learning, which learns a ``diagonal negative combination'' for each query as its explicit semantic opposite-anchor in the embedding space. Finally, we design Boundary-based Targeted Unlearning, which models the noisy correction process as an optimal transport problem, elegantly avoiding Unlearning Backlash. Extensive experiments on benchmark datasets (FashionIQ and CIRR) demonstrate that ConeSep significantly outperforms current state-of-the-art methods, which fully demonstrates the effectiveness and robustness of our method.UniCVR: From Alignment to Reranking for Unified Zero-Shot Composed Visual Retrieval
Air-Know: Arbiter-Calibrated Knowledge-Internalizing Robust Network for Composed Image Retrieval
INTENT: Invariance and Discrimination-aware Noise Mitigation for Robust Composed Image Retrieval
HABIT: Chrono-Synergia Robust Progressive Learning Framework for Composed Image Retrieval
Sketch and Text Synergy: Fusing Structural Contours and Descriptive Attributes for Fine-Grained Image Retrieval
G-MIXER: Geodesic Mixup-based Implicit Semantic Expansion and Explicit Semantic Re-ranking for Zero-Shot Composed Image Retrieval
A Sanity Check on Composed Image Retrieval
Geometry-Aware CLIP Retrieval via Local Cross-Modal Alignment and Steering
FashionMV: Product-Level Composed Image Retrieval with Multi-View Fashion Data
FIRE-CIR: Fine-grained Reasoning for Composed Fashion Image Retrieval
WRF4CIR: Weight-Regularized Fine-Tuning Network for Composed Image Retrieval
Beyond Semantic Search: Towards Referential Anchoring in Composed Image Retrieval
MELT: Improve Composed Image Retrieval via the Modification Frequentation-Rarity Balance Network
modifying the reference image according to the text instructions''. However, existing CIR methods face two limitations: (1) frequency bias leading to Rare Sample Neglect'', and (2) susceptibility of similarity scores to interference from hard negative samples and noise. To address these limitations, we confront two key challenges: asymmetric rare semantic localization and robust similarity estimation under hard negative samples. To solve these challenges, we propose the Modification frEquentation-rarity baLance neTwork MELT. MELT assigns increased attention to rare modification semantics in multimodal contexts while applying diffusion-based denoising to hard negative samples with high similarity scores, enhancing multimodal fusion and matching. Extensive experiments on two CIR benchmarks validate the superior performance of MELT. Codes are available at https://github.com/luckylittlezhi/MELT.