HydraPlus-Net: Attentive Deep Features for Pedestrian Analysis

By Xihui Liu, Haiyu Zhao, Maoqing Tian, Lu Sheng, Jing Shao, Shuai Yi, Junjie Yan, Xiaogang Wang. IEEE International Conference on Computer Vision, 2017

562 citations recorded by the source corpus.

Pedestrian analysis plays a vital role in intelligent video surveillance and is a key component for security-centric computer vision systems. Despite that the convolutional neural networks are remarkable in learning discriminative features from images, the learning of comprehensive features of pedestrians for fine-grained tasks remains an open problem. In this study, we propose a new attentionbased deep neural network, named as HydraPlus-Net (HPnet), that multi-directionally feeds the multi-level attention maps to different feature layers. The attentive deep features learned from the proposed HP-net bring unique advantages: (1) the model is capable of capturing multiple attentions from low-level to semantic-level, and (2) it explores the multi-scale selectiveness of attentive features to enrich the final feature representations for a pedestrian image. We demonstrate the effectiveness and generality of the proposed HP-net for pedestrian analysis on two tasks, i.e. pedestrian attribute recognition and person reidentification. Intensive experimental results have been provided to prove that the HP-net outperforms the state-of-theart methods on various datasets.

Subfields: Artificial Intelligence, Computer Vision and Pattern Recognition

View Paper: https://www.semanticscholar.org/paper/f41c7bb02fc97d5fb9cadd7a49c3e558a1c58a44

How researchers see this paper

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

In Artificial Intelligence

In Computer Vision and Pattern Recognition

Citations in the corpus

Part-Aligned Bilinear Representations for Person Re-identification (2018, background): “Some works [18, 50] proposed the use of attention maps, which are expected to attend to informative body parts.”
Horizontal Pyramid Matching for Person Re-identification (2018, methodology): “Second, several other works abandon the semantic cues for partition[34, 22, 38, 19].”
AlignedReID: Surpassing Human-Level Performance in Person Re-Identification (2017, methodology): “On CUHK03, without re-ranking, HydraPlus-Net [20] achieves 91.”
Multi-level Factorisation Net for Person Re-identification (2018, background): “The others exploited the idea of multi-level fusion either in the form of attention maps computed at multiple intermediate layers of a network [29], or multiple body parts grouped into different levels [53].”
Harmonious Attention Network for Person Re-identification (2018, unknown): “While soft re-id attention modelling is considered in [24], this model assumes tight person boxes thus less suitable for poor detections.”
Beyond Part Models: Person Retrieval with Refined Part Pooling (2017, background): “The latest state of the art on re-∗ [31], PDC [27], DPL [35], Hydra-plus [22] and PAR [37], respectively, which are cropped from the corresponding papers.”
Interaction-And-Aggregation Network for Person Re-Identification (2019, unknown): “Person reID methods focus on two key points: learning a powerful feature representation for images [8, 43, 14, 44, 24, 26, 28, 42, 1] and designing an effective distance metric [32, 29, 56, 2, 57].”
Pose-Guided Feature Alignment for Occluded Person Re-Identification (2019, methodology): “However, the methods proposed in those previous works [31, 13, 32, 38, 24, 42] did not consider a situation that the target person is occluded by various obstructions like cars, trees, or other people.”
Dynamic Dual-Attentive Aggregation Learning for Visible-Infrared Person Re-Identification (2020, background): “Some work [22,26,41] has also investigated using multi-scale or different convolutional channels to capture the pixel-level/small-region-level attentions [35,36].”
Transferable Joint Attribute-Identity Deep Learning for Unsupervised Person Re-identification (2018, methodology): “d process. Our experiments show that the proposed method signiﬁ- cantly outperforms existing models even by using less supervision in the source domain. Attribute for Re-ID Visual semantic attributes [54,36] have been exploited as a mid-level feature representation for cross-view re-id [26,24,25,47,”
Pose-Normalized Image Generation for Person Re-identification (2017, background): “Note that both Spindle [57] and HP-net [29] reported higher results on CUHK03.”
Omni-Scale Feature Learning for Person Re-Identification (2019, methodology): “Dataset and Settings We use PA-100K [28], the largest person attribute recognition dataset.”
Learning Discriminative Features with Multiple Granularities for Person Re-Identification (2018, methodology): “…with strong structural information such as empirical knowledge about human bodies [8, 21, 36, 43] or strong learning-based pose information [33, 44]; 2) Locating part regions by region proposal methods [19, 41]; 3) Enhancing features by middle-level attention on salient partitions [22, 24, 25, 45].”
Human Semantic Parsing for Person Re-identification (2018, unknown): “7 – – HydraPlus [27] – 91.”
Diversity Regularized Spatiotemporal Attention for Video-Based Person Re-identification (2018, background): “[30] proposed a multi-directional attention module to exploit the global and local contents for image-based person re-identification.”
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.”
Pyramidal Person Re-IDentification via Multi-Loss Dynamic Training (2018, methodology): “Inspired by attention models, in [16, 14, 21], the attention-based deep neural networks are proposed to capture multiple attentions and select multi-scale attentive features.”