AlignedReID: Surpassing Human-Level Performance in Person Re-Identification

By Xuan Zhang, Haowen Luo, Xing Fan, Weilai Xiang, Yixiao Sun, Qiqi Xiao, Wei Jiang, Chi Zhang. 2017

544 citations recorded by the source corpus.

In this paper, we propose a novel method called AlignedReID that extracts a global feature which is jointly learned with local features. Global feature learning benefits greatly from local feature learning, which performs an alignment/matching by calculating the shortest path between two sets of local features, without requiring extra supervision. After the joint learning, we only keep the global feature to compute the similarities between images. Our method achieves rank-1 accuracy of 94.4% on Market1501 and 97.8% on CUHK03, outperforming state-of-the-art methods by a large margin. We also evaluate human-level performance and demonstrate that our method is the first to surpass human-level performance on Market1501 and CUHK03, two widely used Person ReID datasets.

Subfields: Biomedical Engineering, Computer Vision and Pattern Recognition

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Biomedical Engineering

1. Tao A. Xiang
2. Kaiyang Zhou
3. Cuiling Lan
4. Wenjun Zeng
5. Shengjin Wang
6. Yifan Sun
7. Yi-Zhe Song
8. Mang Ye

In Computer Vision and Pattern Recognition

1. Tao A. Xiang
2. Luc Van Gool
3. Yi Yang
4. Ling Shao
5. Dacheng Tao
6. Xiaogang Wang
7. Ming-Hsuan Yang
8. Yu Qiao

Citations in the corpus

• Knowledge Transfer Graph for Deep Collaborative Learning (2019, methodology): “DML has been applied to large scale distributed training [30] and re-identification [31].”
• Relation-Aware Global Attention for Person Re-Identification (2019, methodology): “Following the common practices in reid [41, 1, 43], we take ResNet-50 [15] to build our baseline network and integrate our RGA modules into the RetNet50 backbone for effectiveness validation.”
• Perceive Where to Focus: Learning Visibility-Aware Part-Level Features for Partial Person Re-Identification (2019, result): “This observation is consistent with prior works [9, 38, 30, 29] on holistic person re-ID task.”
• Mancs: A Multi-task Attentional Network with Curriculum Sampling for Person Re-Identification (2018, methodology): “It is widely used for learning image embeddings, such as [3, 4, 6, 28, 36, 41].”
• Pose-Normalized Image Generation for Person Re-identification (2017, result): “In adjunct to the work [72] and on Market-1501 dataset, our results (R-1: 95.”
• Cross-Domain Adversarial Feature Learning for Sketch Re-identification (2018, methodology): “[33] aligned persons by dynamically matching the local part features from top to bottom with the minimum total distance.”
• Omni-Scale Feature Learning for Person Re-Identification (2019, methodology): “To exploit the upright body pose, [47, 71, 9, 55] add auxiliary supervision signals to features pooled horizontally from the last convolutional feature maps.”
• Learning to Reduce Dual-Level Discrepancy for Infrared-Visible Person Re-Identification (2019, background): “Existing state-of-the-art singlemodality re-ID methods are very effective in reducing the appearance discrepancy as their retrieval accuracy has already surpassed the accuracy of human [24].”
• Identity-Guided Human Semantic Parsing for Person Re-Identification (2020, methodology): “Some researchers directly partition the person image into rigid horizontal stripes to learn local features [38, 43, 52, 55].”
• Knowledge Distillation for End-to-End Person Search (2019, unknown): “More recent re-identification approaches are based on CNNs [1, 31, 62, 64] and mainly concerned with the estimation of a ID-feature embedding spaces, either via Siamese networks and contrastive losses [1, 9, 11, 15, 31, 36, 50, 56, 62], or with ID-classification and cross-entropy losses [54, 67].”
• Learning Discriminative Features with Multiple Granularities for Person Re-Identification (2018, methodology): “The backbone of our network is ResNet-50 which helps to achieve competitive performances in some Re-ID systems [3, 36, 43].”
• Deep Learning for Person Re-Identification: A Survey and Outlook (2020, methodology): “First, with the advancement of deep learning, most of the image-based Re-ID methods have achieved higher rank-1 accuracy than humans (93.5% [173]) on the widely used Market-1501 dataset.”
• Dynamic Dual-Attentive Aggregation Learning for Visible-Infrared Person Re-Identification (2020, background): “Existing works have achieved human-level performance on the widely-used datasets with end-to-end deep learning [1,14,15,17,39,54], either by global [17,64] or part-level feature learning [40,39,67].”