FashionLens: Toward Versatile Fashion Image Retrieval via Task-Adaptive Learning

Haokun Wen, Xuemeng Song, Xinghao Xie, Xiaolin Chen, Xiangyu Zhao, Weili Guan

arXiv:2605.22552 · 2026-05-22 공개 · arXiv · PDF

mllm e-commerce task-adaptive-learning u-fire fashion-retrieval spherical-query-calibration gradient-guided-sampling image-retrieval

Abstract

Fashion image retrieval is a cornerstone of modern e-commerce systems. A unified framework that supports diverse query formats and search intentions is highly desired in practice. However, existing approaches focus on narrow retrieval tasks and do not fully capture such diversity. Therefore, in this work, we aim to develop a unified framework capable of handling diverse realistic fashion retrieval scenarios, achieving truly versatile fashion image retrieval. To establish a data foundation, we first introduce U-FIRE, a comprehensive benchmark that consolidates fragmented fashion datasets into a unified collection, supplemented by two manually curated datasets for testing generalization. Building upon this, we propose FashionLens, a unified framework based on Multimodal Large Language Models. To handle divergent matching objectives, we design a Proposal-Guided Spherical Query Calibrator that dynamically shifts query representations into task-aligned metric spaces via adaptive spherical linear interpolation. Additionally, to mitigate the optimization imbalance caused by varying task complexities and data scales, we develop a Gradient-Guided Adaptive Sampling strategy that automatically re-weights tasks based on realtime learning difficulty and the data scale prior. Experiments on U-FIRE show that FashionLens achieves state-of-the-art performance across diverse retrieval scenarios and generalizes robustly to unseen tasks. The data and code are publicly released at https://github.com/haokunwen/FashionLens.

한국어 요약

📋 한 줄 요약

**[Fashion Retrieval / Unified MLLM Framework]** FashionLens가 U-FIRE 통합 벤치마크와 Proposal-Guided Spherical Query Calibrator·Gradient-Guided Adaptive Sampling으로 다양한 패션 검색 시나리오 통합 처리, 미관측 task로 강건 일반화.

🎯 핵심 기여도

💡 핵심 아이디어

패션 검색은 task별 narrow model 대신 MLLM 기반 unified framework가 필요하며, query를 task-aligned metric space로 spherical 보간하여 동적으로 shift하고 task 복잡도·데이터 규모에 따라 gradient 기반 자동 재가중하면 다양한 검색 시나리오를 단일 모델로 처리 가능하다.

🔬 기술적 접근법

📊 주요 결과

💭 의의 및 한계

**의의**: 단편 패션 검색 연구의 통합 testbed·MLLM-driven versatile framework 제시, spherical interpolation의 query 적응이 일반화 가능한 패턴, gradient-guided sampling으로 multi-task 최적화 imbalance 정량 완화. **한계**: MLLM backbone 추론 cost, U-FIRE 외 산업 데이터에의 일반화는 추가 검증, fashion-specific 가정의 다른 도메인 이식성 미평가.

🚀 실용적 활용