Auto-ReID+: Searching for a multi-branch ConvNet for person re-identification

Gu, Hongyang; Fu, Guohong; Li, Jianmin; Zhu, Jun

doi:10.1016/j.neucom.2020.12.105

Cited by 13 publications

(2 citation statements)

References 9 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…We compared DBA-Net with the state-of-the-art (SOTA) approaches: AGW [1], CAM [4], ABD-Net [6], BagTricks [9], AANet [10], RAG-SC [12], SGSC [14], GRLL [16], MGN [20], GD-Net [21], IANet [23], Pyramid [24], SONA [27], SCAL [28], Auto-ReID+ [44] and Ms-Mb [45]. We did not adopt the re-ranking trick for the sake of fairness.…”

Section: Comparison With State-of-the-art Methodsmentioning

confidence: 99%

Dual Branch Attention Network for Person Re-Identification

Fan

Wang

Cheng

et al. 2021

Sensors

View full text Add to dashboard Cite

As a sub-direction of image retrieval, person re-identification (Re-ID) is usually used to solve the security problem of cross camera tracking and monitoring. A growing number of shopping centers have recently attempted to apply Re-ID technology. One of the development trends of related algorithms is using an attention mechanism to capture global and local features. We notice that these algorithms have apparent limitations. They only focus on the most salient features without considering certain detailed features. People’s clothes, bags and even shoes are of great help to distinguish pedestrians. We notice that global features usually cover these important local features. Therefore, we propose a dual branch network based on a multi-scale attention mechanism. This network can capture apparent global features and inconspicuous local features of pedestrian images. Specifically, we design a dual branch attention network (DBA-Net) for better performance. These two branches can optimize the extracted features of different depths at the same time. We also design an effective block (called channel, position and spatial-wise attention (CPSA)), which can capture key fine-grained information, such as bags and shoes. Furthermore, based on ID loss, we use complementary triplet loss and adaptive weighted rank list loss (WRLL) on each branch during the training process. DBA-Net can not only learn semantic context information of the channel, position, and spatial dimensions but can integrate detailed semantic information by learning the dependency relationships between features. Extensive experiments on three widely used open-source datasets proved that DBA-Net clearly yielded overall state-of-the-art performance. Particularly on the CUHK03 dataset, the mean average precision (mAP) of DBA-Net achieved 83.2%.

show abstract

Section: Comparison With State-of-the-art Methodsmentioning

confidence: 99%

Dual Branch Attention Network for Person Re-Identification

Fan

Wang

Cheng

et al. 2021

Sensors

View full text Add to dashboard Cite

show abstract

“…The global correlative network (GCN) [4] was introduced as a solution to leverage the combination of local and global features, effectively elevating the quality of discriminative representations. An automatic search for a multi-branch CNN tailored to person ReID tasks was introduced in [5], employing neural architecture search (NAS), and multi-resolution, multi-branch macro search architecture were utilized to enhance scale information extraction. To address the issue of insufficient diversity in local features, a network named ABDNet was designed in [6].…”

Section: Introductionmentioning

confidence: 99%

Foreground-aware Transformer Network for Person Re-identification

Zhang,

Tan,

Fang

2023

Preprint

View full text Add to dashboard Cite

Transformer-based person re-identification (person ReID) technologies tend to capture global information, but only focus on global features and ignore the interference of irrelevant information. To the best of our knowledge, most foreground information corresponds to pedestrian(in the person ReID datasets), enhancing foreground information or weakening background information helps to distinguish the person from the background. From this insight, we proposed a foreground-aware transformer network to achieve the task of person ReID. To make the most of foreground information for person identification, we isolate the foreground by minimizing the impact of background interference and introduce a foreground-aware loss function. This loss function directs the attention of networks toward the primary foreground information in the image, optimizing its ability to identify pedestrian. To prove the effectiveness of our proposed foreground-aware transformer network, we conducted experiments on Market1501 and MSMT17 datasets. Our experimental results indicate that the proposed method can yield substantial improvements in person ReID accuracy, demonstrating the practical value of our foreground-aware transformer network in addressing real-world person ReID challenges.

show abstract