Dynamic Dual-Attentive Aggregation Learning for Visible-Infrared Person Re-identification

Ye, Mang; Shen, Jianbing; Crandall, David J.; Shao, Ling; Luo, Jiebo

doi:10.1007/978-3-030-58520-4_14

Cited by 241 publications

(136 citation statements)

References 44 publications

Supporting

Mentioning

109

Contrasting

Order By: Relevance

“…We need to measure this degree of importance effectively. For example, the IWPA method [52] and the CPA method [53] extract the weights of the features after the partition, which has a sure consistency with the scheme proposed in this paper. However, these two schemes are still quite different from the CPWA method we proposed.…”

Section: Related Work a Person Re-identification Methods Of Part-basedmentioning

confidence: 57%

“…After obtaining the weights between each partition and each channel, our method is fused with the input feature map so that the output of our CPWA module is a feature map that has precisely the same dimensions as the input feature map so that it can be embedded in any convolutional layer in the network structure. CPA [53] is based on IWPA [52] by adding reshape and upsample operations so that the CPA module, like our CPWA module, can be easily embedded in many layers of the network structure. However, calculating the weight between the local regions uses a series of convolutions to generate three new intermediate feature maps.…”

Section: Related Work a Person Re-identification Methods Of Part-basedmentioning

confidence: 99%

“…The other The associate editor coordinating the review of this manuscript and approving it for publication was G. R. Sinha . is a type of method that divides the pedestrian image into different partitions, extracts the features of each area separately, and then merges them together [12], [13], [52], [53]. This type of method has different schemes for partitioning pedestrian images.…”

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

Person Re-Identification Based on Partition Adaptive Network Structure and Channel Partition Weight Adaptive

Chen¹,

Yang

2021

IEEE Access

View full text Add to dashboard Cite

This article noticed that the feature of the block in human body image is noticeable, some algorithms in pedestrian re-identification are based on partitioning the human body to calculate the similarity between pedestrians images. However, it is not easy to find a proper method to partition the pedestrian image. This article proposes a method partition adaptive network structure (PANS) to automatically determine the partition scheme. This method can save much work to find a suitable partitioning scheme, and the effect of the automatic partitioning scheme is better than that of the manual partitioning scheme. In addition, this paper proposes a pedestrian re-recognition method based on channel partition weight adaptive (CPWA). We obtained better results on three public pedestrian re-identification data sets compared with the baseline. This method, combined with the automatic partitioning scheme proposed in this article, can improve results. We have done experiments on the three public data sets of market-1501, DukeMTMC-Reid, and CUHK03, proving the superiority of the two methods proposed in this article.

show abstract

Section: Related Work a Person Re-identification Methods Of Part-basedmentioning

confidence: 57%

Section: Related Work a Person Re-identification Methods Of Part-basedmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Person Re-Identification Based on Partition Adaptive Network Structure and Channel Partition Weight Adaptive

Chen¹,

Yang

2021

IEEE Access

View full text Add to dashboard Cite

show abstract

“…ST-GCN [36] explored both spatial and temporal information via GCN for skeleton-based action recognition. Compared with DDAG [37] which utilizes a graph-based attention module to aggregate the cross-modality part-level feature, A 2 G refines the feature representation via a graph neural network on the pedestrian attribute graph to further explore the similarity of attribute space. In this paper, we construct a pedestrian attribute graph and aggregate the features through graph embedding algorithm, which is a novel approach for attributeauxiliary person re-ID.…”

Section: Graph Representation Learning On Computer Vision Tasksmentioning

confidence: 99%

Graph Neural Network Based Attribute Auxiliary Structured Grouping for Person Re-Identification

Tang

Gao²,

Chen

et al. 2024

IEEE Access

View full text Add to dashboard Cite

Recently, person re-identification (re-ID) with weakly labeled or unlabeled data has drawn much attention in open-set and cross-domain re-ID systems especially for the attribute auxiliary weakly supervised person re-ID. Although state-of-the-art clustering-based methods have achieved good performance, the pseudo labels generated through clustering are often low-quality and noisy. To address this problem, we propose a graph neural network based Attribute Auxiliary structured Grouping (A 2 G) to improve the confidence of the pseudo labels. Different from the existing clustering-based approaches that only utilize the similarity in feature space, we learn the feature representation from the similarities in both attribute space and feature space by graph learning on the pedestrian attribute graph. Specifically, we first utilize the pair-wise attribute similarity to represent the relation between two pedestrians to construct a pedestrian attribute graph. Furthermore, we aggregate the features from their neighborhood on a pedestrian attribute graph by the graph neural network, which would make the attribute similar pairs closer and simultaneously take the dissimilar pairs further in the feature space. Finally, to avoid the over-confidence of the hard pseudo labels, we regularize the learning of the embedding model with the smoothed pseudo label (SPL) in the optimization of the feature embedding network. We conduct extensive experiments on several person re-ID datasets to validate the efficacy of our proposed method. The results demonstrate that our technique is effective and promising for person re-ID tasks.INDEX TERMS Unsupervised person re-identification, attribute-auxiliary structured grouping, graph neural network.

show abstract

“…In each training batch, we first randomly select N identities, then select N K visible images of N identities as input to the RGB branch of our network, and N K infrared images to send to the IR branch (N = K = 4 in our experiment). The above sampling method is similar to [13] for identity balance. Then, we take the output feature maps of each modality-shared layer as the input to the corresponding DDSL module.…”

Section: B Dual-path Deep Supervision Learningmentioning

confidence: 99%

Dual-Path Deep Supervision Network with Self-Attention for Visible-Infrared Person Re-Identification

Cheng

Xiao

et al. 2021

2021 IEEE International Symposium on Circuits and Systems (ISCAS)

View full text Add to dashboard Cite

Visible-infrared person re-identification(VI-ReID) is an emerging but challenging problem that aims to match pedestrians captured by visible and infrared cameras. Existing studies in this field mainly focus on learning sharable feature representations from the last layer of deep convolution neural networks (CNNs) to handle the cross-modality discrepancies. However, due to the huge differences between visible and infrared images, the last layer's feature representations are less discriminative for VI-ReID. To remedy this, we propose a novel deep supervision learning network, namely Dual-path Deep Supervision Network (DDSN), for VI-ReID. Based on the backbone network, DDSN consists of two key modules, (1) a dual-path deep supervision learning (DDSL) module that is plugged into multiple network layers, and (2) a self-attention module that is developed on top of the backbone network. The backbone network extracts multilevel features at lower middle layers, and several DDSL modules utilize these features to generate more discriminative descriptors. Furthermore, we apply the self-attention module for context modeling to capture useful contextual cues as a supplement. By fusing these descriptors, DDSN can utilize both multi-level information and potential contextual information. Despite the apparent simplification, our method outperforms several state-ofthe-art methods on two large-scale datasets: RegDB and SYSU-MM01.

show abstract

Dynamic Dual-Attentive Aggregation Learning for Visible-Infrared Person Re-identification

Cited by 241 publications

References 44 publications

Person Re-Identification Based on Partition Adaptive Network Structure and Channel Partition Weight Adaptive

Person Re-Identification Based on Partition Adaptive Network Structure and Channel Partition Weight Adaptive

Graph Neural Network Based Attribute Auxiliary Structured Grouping for Person Re-Identification

Dual-Path Deep Supervision Network with Self-Attention for Visible-Infrared Person Re-Identification

Contact Info

Product

Resources

About