Street-to-Shop: Cross-Scenario Clothing Retrieval via Parts Alignment and Auxiliary Set

1. Paper is from national university of singapore, cvpr, 2012

2. 主要内容：In this paper, we address a practical problem of crossscenario clothing retrieval - given a daily human photo captured in general environment, e.g., on street, finding similar clothing in online shops, where the photos are captured more professionally and with clean background. There are large discrepancies between daily photo scenario and online shopping scenario.就是在日常环境下的图像，如何查询到相似的产品。

步骤：

首先，通过训练human detector来定位30个human parts；

然后，通过两步计算，获得更加可靠的 one-to-many的查询图像和在线shopping 图像之间相似性的计算。 

1） 通过直接的sparse reconstruction得到auxiliary set；

2） 通过跨场景的多对多的相似性传递矩阵（由线下的额外的辅助集和在线shopping set） 

本论文更加注重于训练。