Re-ID论文 Person Search by Multi-Scale Matching

We consider the problem of person search in unconstrained scene images.
 Existing methods usually focus on improving the person detection accuracy to mitigate negative effects
 imposed by misalignment, misdetections, and false alarms resulted from noisy people autodetection. 
In contrast to previous studies, we show that sufficiently reliable person instance cropping is achievable by slightly improved state of-the-art deep learning object detectors (e.g. Faster-RCNN), and the under-studied multiscale matching problem in person search is a more severe barrier.
 In this work, we address this multi-scale person search challenge by proposing a Cross-Level Semantic Alignment (CLSA) deep learning approach capable of learning more discriminative identity feature representations in a unified end-to-end model.

 This is realised by exploiting the in-network feature pyramid structure of a deep neural network enhanced by a novel cross pyramid-level semantic alignment loss function.
 This favourably eliminates the need for constructing a computationally expensive image pyramid and a complex multi-branch network architecture. 
Extensive experiments show the modelling advantages and performance superiority of CLSA over the state-of-the-art person search and multiscale matching methods on two large person search benchmarking datasets: CUHK-SYSU and PRW.

研究了无约束场景图像中的人物搜索问题。

现有的检测方法通常侧重于提高人员检测的准确性，以减轻负面影响由噪声人的自动检测导致的失调、误检和误报。

与以往的研究相比，我们发现，通过稍微改进现有的深度学习对象检测器(如fast - rcnn)可以实现足够可靠的person实例裁剪，而在person搜索中尚未得到充分研究的多尺度匹配问题是一个更为严重的障碍。

在这项工作中，我们提出了一种跨级别语义对齐(CLSA)深度学习方法，能够在统一的端到端模型中学习更有区别的身份特征表示，从而解决这个多尺度的人员搜索挑战。

这是通过利用深度神经网络的网络内特征金字塔结构实现的，该结构由一种新的交叉金字塔级语义对齐损失函数增强。这有利于消除构建计算昂贵的图像金字塔和复杂的多分支网络体系结构的需要。

大量实验表明，里昂证券在两个大型人员搜索基准数据集(CUHK-SYSU和PRW)上的建模优势和性能优于最先进的人员搜索和多尺度匹配方法。