Deep Representation Learning With Part Loss for Person Re-Identification

By Hantao Yao, Shiliang Zhang, Richang Hong, Yongdong Zhang, Changsheng Xu, Qi Tian. IEEE Transactions on Image Processing, 2017

433 citations recorded by the source corpus.

Learning discriminative representations for unseen person images is critical for person re-identification (ReID). Most of the current approaches learn deep representations in classification tasks, which essentially minimize the empirical classification risk on the training set. As shown in our experiments, such representations easily get over-fitted on a discriminative human body part on the training set. To gain the discriminative power on unseen person images, we propose a deep representation learning procedure named part loss network, to minimize both the empirical classification risk on training person images and the representation learning risk on unseen person images. The representation learning risk is evaluated by the proposed part loss, which automatically detects human body parts and computes the person classification loss on each part separately. Compared with traditional global classification loss, simultaneously considering part loss enforces the deep network to learn representations for different body parts and gain the discriminative power on unseen persons. Experimental results on three person ReID datasets, i.e., Market1501, CUHK03, and VIPeR, show that our representation outperforms existing deep representations.

Subfields: Computer Vision and Pattern Recognition

View Paper: https://www.semanticscholar.org/paper/a6b163dcca2054df8629cfd66c26700409c39604

How researchers see this paper

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

In Computer Vision and Pattern Recognition

Citations in the corpus

Batch Coherence-Driven Network for Part-Aware Person Re-Identification (2020, background): “Part detection provides the spatial location of body parts, usually in the form of bounding boxes [18], [21], [24], from which the part-level features are extracted.”
Pyramidal Person Re-IDentification via Multi-Loss Dynamic Training (2018, background): “A part loss is proposed in [29], which automatically detects human body parts and computes the person classiﬁcation loss on each part separately.”
Dynamic Dual-Attentive Aggregation Learning for Visible-Infrared Person Re-Identification (2020, background): “These challenges make it difficult to reliably learn discriminative part-level features using state-of-the-art single-modality Re-ID systems [40,45,55,67].”
Deep Metric Learning: A Survey (2019, background): “[76] presented part loss for person re-identification task to take advantage of learning risk minimization constraint, as in SVM.”
AlignedReID: Surpassing Human-Level Performance in Person Re-Identification (2017, background): “Some works consider local information by dividing images into several parts without an alignment [33, 38, 43], but these methods suffer from inaccurate detection boxes, occlusion and pose misalignment.”
Beyond Part Models: Person Retrieval with Refined Part Pooling (2017, background): “First, no matter which partition strategy is applied in PCB, it signiﬁcantly outperforms PAR [37], which learns to partition through attention mechanism.”
Learning Discriminative Features with Multiple Granularities for Person Re-Identification (2018, methodology): “Softmax loss is almost the only choice of classification loss function for its strong robustness to various kinds of multi-class classification tasks, which can be used individually [1, 42, 34, 32, 24, 35, 18, 21] or combined with other losses [19, 7, 37, 3] in embedding learning procedures for Re-I”
ABD-Net: Attentive but Diverse Person Re-Identification (2019, methodology): “[9] proposed a Part Loss Network which defined a loss for each average pooled body part and jointly optimized the summation losses.”
Horizontal Pyramid Matching for Person Re-identification (2018, methodology): “Second, several other works abandon the semantic cues for partition[34, 22, 38, 19].”
GaitPart: Temporal Part-Based Model for Gait Recognition (2020, background): “More evidences from [6, 15, 20, 5, 12, 13, 19, 28] have implied that partial features for human body description can offer ﬁne-grained information, which is conducive to individual identiﬁcation task.”
AANet: Attribute Attention Network for Person Re-Identifications (2019, background): “[24] proposed the Part Loss Networks (PL-Net) to automatically detect human parts and cross train them with the main identity task.”
Pose-Normalized Image Generation for Person Re-identification (2017, background): “A number of models [39, 62, 58, 57, 23, 50, 45] are proposed to address this problem.”
Dual Attention Matching Network for Context-Aware Feature Sequence Based Person Re-identification (2018, methodology): “…correspondence structure between local patterns but also automatically remove local corrup-tions within sequence, our method achieves better performance than body-part based ( e.g ., SpindleNet [59], DRL-PL [52]) and densely-matching based ( e.g ., SCSP [3]) meth-ods on dataset Market-1501.”
Pose-Driven Deep Convolutional Model for Person Re-identification (2017, background): “Therefore, many researchers attempt to leverage deep models in person ReID [1, 10, 53, 46, 42, 61, 13, 56, 43, 57].”
Cross-Domain Adversarial Feature Learning for Sketch Re-identification (2018, background): “Some researchers learnt local features frommultiple body parts of persons [26, 27], demonstrating more effective than dividing person images into several fixed-length strips [4].”
Deep Learning for Person Re-Identification: A Survey and Outlook (2020, background): “2), it aggregates part-level local features to formulate a combined representation for each person image [66], [67], [68]; c) Auxiliary Feature (§ 2.”
Perceive Where to Focus: Learning Visibility-Aware Part-Level Features for Partial Person Re-Identification (2019, background): “Recent works [23, 12, 26, 29, 22, 28, 16] further advance the state of the art on holistic person re-ID, through learning part-level deep features.”
Pseudo-Labeling and Confirmation Bias in Deep Semi-Supervised Learning (2019, background): “Convolutional neural networks (CNNs) have become the dominant approach for many computer vision tasks [8, 17, 2, 18, 38, 11, 37].”