Zhedong Zheng

Zhedong Zheng

Asst. Prof. @ UMacau
IEEE Senior Member

A discriminatively learned CNN embedding for person reidentification

Authors: Zhedong Zheng, liang-zhengLiang Zheng, yi-yangYi Yang

Published in ACM Transactions on Multimedia Computing, Communications, and Applications (TOMM), 2018

Recommended citation: Zhedong Zheng, Liang Zheng, Yi Yang, "A discriminatively learned CNN embedding for person reidentification." ACM Transactions on Multimedia Computing, Communications, and Applications (TOMM), 2018. DOI: 10.1145/3159171
Download PDF: https://zdzheng.xyz/files/2018/TOMM18.pdf

Code is available at: https://github.com/layumi/2016_person_re-ID

Abstract: In this article, we revisit two popular convolutional neural networks in person re-identification (re-ID): verification and identification models. The two models have their respective advantages and limitations due to different loss functions. Here, we shed light on how to combine the two models to learn more discriminative pedestrian descriptors. Specifically, we propose a Siamese network that simultaneously computes the identification loss and verification loss. Given a pair of training images, the network predicts the identities of the two input images and whether they belong to the same identity. Our network learns a discriminative embedding and a similarity measurement at the same time, thus taking full usage of the re-ID annotations. Our method can be easily applied on different pretrained networks. Albeit simple, the learned embedding improves the state-of-the-art performance on two public person re-ID benchmarks. Further, we show that our architecture can also be applied to image retrieval.

Keywords:

embedding person reidentification cnn embedding person discriminatively learned cnn object re-identification content-based retrieval person re-id person retrieval person search 
@article{zheng2018discriminatively,
author = "Zheng, Zhedong and Zheng, Liang and Yang, Yi",
doi = "10.1145/3159171",
title = "A discriminatively learned CNN embedding for person reidentification",
abstract = "In this article, we revisit two popular convolutional neural networks in person re-identification (re-ID): verification and identification models. The two models have their respective advantages and limitations due to different loss functions. Here, we shed light on how to combine the two models to learn more discriminative pedestrian descriptors. Specifically, we propose a Siamese network that simultaneously computes the identification loss and verification loss. Given a pair of training images, the network predicts the identities of the two input images and whether they belong to the same identity. Our network learns a discriminative embedding and a similarity measurement at the same time, thus taking full usage of the re-ID annotations. Our method can be easily applied on different pretrained networks. Albeit simple, the learned embedding improves the state-of-the-art performance on two public person re-ID benchmarks. Further, we show that our architecture can also be applied to image retrieval.",
journal = "ACM Transactions on Multimedia Computing, Communications, and Applications (TOMM)",
volume = "14",
number = "1",
pages = "13",
year = "2018",
publisher = "ACM",
code = "https://github.com/layumi/2016\_person\_re-ID",
url = "https://zdzheng.xyz/files/2018/TOMM18.pdf" }