A Discriminatively Learned CNN Embedding for Person Reidentification

2016/11/30 by Zhedong Zheng, Liang Zheng, Yi Yang · 7 citations
Computer Science · #Artificial intelligence #Code (set theory) #Computer science #Convolutional neural network #Discriminative model #Embedding #Face recognition and analysis #Human Pose and Action Recognition #Identification (biology) #Identity (music) #Image (mathematics) #Machine learning #Pattern recognition (psychology) #Similarity (geometry) #Simple (philosophy) #Video Surveillance and Tracking Methods #cs.CV

paper · pdf · doi:10.1145/3159171

arxiv created 2017/02/03 · openalex publication_date 2017/12/13 · arxiv updated 2018/04/04 · openalex created_date 2025/10/10 · openalex updated_date 2026/08/05

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

Citations

• Very Deep Convolutional Networks for Large-Scale Image Recognition
• Long Short-Term Memory
• Deep Metric Learning for Practical Person Re-Identification
• ImageNet Large Scale Visual Recognition Challenge
• DeepID3: Face Recognition with Very Deep Neural Networks
• Deep Learning Face Representation by Joint Identification-Verification
• ImageNet classification with deep convolutional neural networks
• Particular object retrieval with integral max-pooling of CNN activations
• Person Re-identification in the Wild
• Person Re-identification: Past, Present and Future
• Large-Margin Softmax Loss for Convolutional Neural Networks
• Learning a Discriminative Null Space for Person Re-identification
• SIFT Meets CNN: A Decade Survey of Instance Retrieval
• MatConvNet - Convolutional Neural Networks for MATLAB
• Deep Transfer Learning for Person Re-identification
• Multiregion Bilinear Convolutional Neural Networks for Person Re-Identification

Cited by

• A heterogeneous branch and multi-level classification network for person re-identification
• Grafted network for person re-identification
• GLAD
• Person Re-Identification by Semantic Region Representation and Topology Constraint
• An Evaluation of Deep CNN Baselines for Scene-Independent Person Re-identification
• Multi-Task Learning With Coarse Priors for Robust Part-Aware Person Re-Identification
• PGGANet: Pose Guided Graph Attention Network for Person Re-identification