Density-Adaptive Kernel based Re-Ranking for Person Re-Identification

Guo, Ruopei; Li, Chunguang; Li, Yonghua; Jiang, Lin

doi:10.1109/icpr.2018.8545619

Cited by 7 publications

(6 citation statements)

References 42 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Moreover, our model has a low complexity for its operation on FC layers and is compatible with the strong backbone (ResNet50). In addition, compared to a traditional reranking method such as [13], which only refines the ranking list by the context information of the testing set, our model is more stable and stronger for learning more discriminative information from the training set.…”

Section: Comparison To State-of-the-art Resultsmentioning

confidence: 99%

“…Many works [12], [13], [21]- [26] have reported that the original ranking list can be refined by extra contextual information in the gallery set. For example, in [21], [22], extra information is obtained from human feedback; in [12], [13], [23]- [26], the contextual information in the local neighborhood structure or the manifold structure of the gallery samples were considered. Additionally, a unified framework is proposed in [27], which utilizes both neighbor information and user feedback to refine the ranking scores for cross-modal retrieval.…”

Section: B Rerankingmentioning

confidence: 99%

“…Nevertheless, these methods suffer from high computation costs and are hardly suitable for the datasets of large size. In [11]- [13], reranking as a post-processing approach is proposed to refine the ranking list based on the contextual information of the probe image and the gallery set. However, the improvement of reranking is usually restrained because exploitable information in reranking methods is very limited.…”

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

Deep Group-Shuffling Dual Random Walks With Label Smoothing for Person Reidentification

Guo

Lin

et al. 2020

IEEE Access

Self Cite

View full text Add to dashboard Cite

Person reidentification (ReID) is a challenging task of finding a target pedestrian in a gallery set collected from multiple nonoverlapping camera views. Recently, state-of-the-art ReID performance has been achieved via an end-to-end trainable deep neural network framework, which integrates convolution feature extraction, similarity learning and reranking into a joint optimization framework. In such a framework, the similarity is learned via an embedding network, the reranking is conducted with a random walk, and the whole framework is optimized with a cross-entropy-based verification loss. Unfortunately, the embedding net is difficult to train well because their two-dimensional outputs mutually interfere each other when using the conventional random walk. In addition, the supervision information has not been fully exploited during the training phase due to the binary nature of the verification loss. In this paper, we propose a novel approach, called group-shuffling dual random walks with label smoothing (GSDRWLS), in which random walks are performed separately on two channels-one for positive verification and one for negative verificationand the binary verification labels are properly modified with an adaptive label smoothing technique before feeding into the verification loss in order to train the overall network effectively and to avoid the overfitting problem. Extensive experiments conducted on three large benchmark datasets, including CUHK03, Market-1501 and DukeMTMC, confirm the superior performance of our proposal. INDEX TERMS Person reidentification, deep neural network, dual random walks, label smoothing.

show abstract

Section: Comparison To State-of-the-art Resultsmentioning

confidence: 99%

Section: B Rerankingmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Deep Group-Shuffling Dual Random Walks With Label Smoothing for Person Reidentification

Guo

Lin

et al. 2020

IEEE Access

Self Cite

View full text Add to dashboard Cite

show abstract

“…The primary idea behind re-ranking [27][28][29][30][31][32][33][34] is to optimize the initial ranking list by utilizing gallery-to-gallery similarities to boost the retrieval performance at post-processing steps. Guo et al [27] introduce inverse density-adaptive kernel-based re-ranking (inv-DAKR) and bidirectional density-adaptive kernel-based re-ranking (bi-DAKR), two simple yet efficient re-ranking algorithms based on a smooth kernel function with a densityadaptive parameter. Xu et al [28] present a feature-relation-map-based similarity evaluation (FRM-SE) model that uses convolution operations to automatically mine the latent relations between k-neighbors, reducing queries and memory use.…”

Section: Re-ranking For Person Re-idmentioning

confidence: 99%

Swin Transformer Based on Two-Fold Loss and Background Adaptation Re-Ranking for Person Re-Identification

et al. 2022

View full text Add to dashboard Cite

Person re-identification (Re-ID) aims to identify the same pedestrian from a surveillance video in various scenarios. Existing Re-ID models are biased to learn background appearances when there are many background variations in the pedestrian training set. Thus, pedestrians with the same identity will appear with different backgrounds, which interferes with the Re-ID performance. This paper proposes a swin transformer based on two-fold loss (TL-TransNet) to pay more attention to the semantic information of a pedestrian’s body and preserve valuable background information, thereby reducing the interference of corresponding background appearance. TL-TransNet is supervised by two types of losses (i.e., circle loss and instance loss) during the training phase. In the retrieval phase, DeepLabV3+ as a pedestrian background segmentation model is applied to generate body masks in terms of query and gallery set. The background removal results are generated according to the mask and are used to filter out interfering background information. Subsequently, a background adaptation re-ranking is designed to combine the original information with the background-removed information, which digs out more positive samples with large background deviation. Extensive experiments on two public person Re-ID datasets testify that the proposed method achieves competitive robustness performance in terms of the background variation problem.

show abstract

“…Zhong et al [28] combine the Jaccard distance and the original distance, complete re-ranking with k-reciprocal encoding. Guo et al [29] exploit a density-adaptive kernel technique to perform efficient re-ranking for person re-ID.…”

Section: Re-rankingmentioning

confidence: 99%

Contribution-Based Multi-Stream Feature Distance Fusion Method With <inline-formula> <tex-math notation="LaTeX">${k}$ </tex-math> </inline-formula>-Distribution Re-Ranking for Person Re-Identification

Huang

et al. 2019

IEEE Access

View full text Add to dashboard Cite

Person re-identification (re-ID) is an important and challenging topic in video surveillance and public security. Re-ID aims to retrieve persons from different cameras. Despite the developments in recent years, re-ID still faces many challenges due to different camera views, changeable person posture, complex background, and occlusion. To exploit more discriminative image similarity descriptor, we propose a novel method in this paper. First, we design a body partition extraction network to extract three body regions with efficient alignment. Second, we propose a multi-stream contribution framework to fuse feature distance with different contributions and generate the final image similarity descriptor. In addition, we combine re-ID and semantic segmentation. A mask feature is introduced to the proposed framework and we design a contribution feedback module to generate contribution coefficients dynamically. Third, in order to improve re-ID performance, we propose a fragment learning method to optimize the contribution feedback module. Fourth and last, we propose a k-distribution re-ranking strategy to further improve performance. Our method achieves competitive results on two popular datasets, CUHK03, and Market1501, with rank-1 accuracy of 93.5% and 85.7%. The proposed re-ranking method achieves 2.3% and 2.8% performance boost. The data demonstrate the effectiveness of the proposed multi-stream contribution framework and the k-distribution re-ranking strategy.

show abstract

Density-Adaptive Kernel based Re-Ranking for Person Re-Identification

Cited by 7 publications

References 42 publications

Deep Group-Shuffling Dual Random Walks With Label Smoothing for Person Reidentification

Deep Group-Shuffling Dual Random Walks With Label Smoothing for Person Reidentification

Swin Transformer Based on Two-Fold Loss and Background Adaptation Re-Ranking for Person Re-Identification

Contribution-Based Multi-Stream Feature Distance Fusion Method With <inline-formula> <tex-math notation="LaTeX">${k}$ </tex-math> </inline-formula>-Distribution Re-Ranking for Person Re-Identification

Contact Info

Product

Resources

About