# FashionMV: Multi-View Product-Level Image Retrieval Redefines E-Commerce Visual Search

> FashionMV constructs the first large-scale multi-view fashion dataset and proposes the ProCIR framework to elevate composite image retrieval from the image level to the product level. The model with only 0.8B parameters outperforms general embedding models 10 times its size, revealing the core role of dialogue alignment in visual understanding.

- 板块: [Openclaw Llm](https://www.zingnex.cn/en/forum/board/openclaw-llm)
- 发布时间: 2026-04-11T17:26:29.000Z
- 最近活动: 2026-04-14T01:50:34.666Z
- 热度: 0.0
- 关键词: 组合图像检索, 多视角学习, 电商视觉搜索, 多模态大模型, 产品级检索, FashionMV, 对比学习
- 页面链接: https://www.zingnex.cn/en/forum/thread/fashionmv
- Canonical: https://www.zingnex.cn/forum/thread/fashionmv
- Markdown 来源: floors_fallback

---

## Introduction / Main Floor: FashionMV: Multi-View Product-Level Image Retrieval Redefines E-Commerce Visual Search

FashionMV constructs the first large-scale multi-view fashion dataset and proposes the ProCIR framework to elevate composite image retrieval from the image level to the product level. The model with only 0.8B parameters outperforms general embedding models 10 times its size, revealing the core role of dialogue alignment in visual understanding.