Re-ID Driven Localization Refinement for Person Search

Han, Chuchu; Ye, Jiacheng; Zhong, Yong; Tan, Xin; Zhang, Chi; Gao, Changxin; Sang, Nong

doi:10.1109/iccv.2019.00991

Cited by 120 publications

(81 citation statements)

References 34 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…The task-cascaded methods adopt two separate steps of detection and re-ID. In this group, MGTS [3] uses VGGNet [27] based Faster R-CNN [25] detector, CLSA [16] employs vanilla Faster R-CNN, while RDLR [12] uses Faster R-CNN with FPN [19]. Evaluation on PRW.…”

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

confidence: 99%

“…These methods use separate pre-trained pedestrian detectors and only learn the re-ID networks, thus are not end-toend trainable. Recently, Hanet al [12] propose to refine localization using re-ID training loss, in which way person-to-person clutter can be reduced while accessory could be appropriately encompassed for re-ID feature extraction.…”

Section: Related Workmentioning

confidence: 99%

“…The pedestrian detection branch performs bounding boxes binary classification and regression, which are then used to pool Region of Interest (RoI) features from the backbone (a.k.a RoI Pooling [9] in detection) and send it to a separate re-ID branch for identification features extraction. The task-cascaded methods [3,12,16,40] formulate it as a cascaded framework. In the pipeline, a pedestrian detector first takes a full-size video frame as input and produces bounding boxes of pedestrians.…”

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

Bottom-Up Foreground-Aware Feature Fusion for Person Search

Yang

Chen

et al. 2020

Proceedings of the 28th ACM International Conference on Multimedia

View full text Add to dashboard Cite

The key to efficient person search is jointly localizing pedestrians and learning discriminative representation for person re-identification (re-ID). Some recently developed task-joint models are built with separate detection and re-ID branches on top of shared region feature extraction networks, where the large receptive field of neurons leads to background information redundancy for the following re-ID task. Our diagnostic analysis indicates the task-joint model suffers from considerable performance drop when the background is replaced or removed. In this work, we propose a subnet to fuse the bounding box features that pooled from multiple ConvNet stages in a bottom-up manner, termed bottom-up fusion (BUF) network. With a few parameters introduced, BUF leverages the multi-level features with different sizes of receptive fields to mitigate the backgroundbias problem. Moreover, the newly introduced segmentation head generates a foreground probability map as guidance for the network to focus on the foreground regions. The resulting foreground attention module (FAM) enhances the foreground features. Extensive experiments on PRW and CUHK-SYSU validate the effectiveness of the proposals. Our Bottom-Up Foreground-Aware Feature Fusion (BUFF) network achieves considerable gains over the state-of-thearts on PRW and competitive performance on CUHK-SYSU.

show abstract

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

confidence: 99%

Section: Related Workmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Bottom-Up Foreground-Aware Feature Fusion for Person Search

Yang

Chen

et al. 2020

Proceedings of the 28th ACM International Conference on Multimedia

View full text Add to dashboard Cite

show abstract

“…Some two-step methods [1,2,3] improved the feature extraction and metric learning of the re-identification model. Some two-step methods [4,5,6] had improvements on the correlation between the detection model and the re-identification model.…”

Section: Introductionmentioning

confidence: 99%

Stable and Effective One-Step Method for Person Search

Xiang

Wang

et al. 2021

ICASSP 2021 - 2021 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP)

View full text Add to dashboard Cite

Person search, which requires both pedestrian detection and person re-identification, is a challenging computer vision task applied to real-world scenarios. The challenges faced by detection and re-identification, such as occlusion, poor illumination, confusing background, are still urgent for person search. In addition, one-step methods for person search need to deal with the divergence between two tasks. In this work, we propose an end-to-end model containing the feature extractor, the region proposal network, and the multi-task learning module. In order to process divergence between detection and re-identification, we introduce switchable normalization and gradient centralization to improve the stability of the model. To solve the imbalance problem of hard examples, we introduce focal loss as a classification loss in the multi-task learning module. The experimental results on two benchmarks, i.e., CUHK-SYSU and PRW, well demonstrate that our method outperforms the state-of-the-art one-step methods.

show abstract

“…由于训练数据与测试数据的行人身份不重合, 所以也看做一种零样本学习 [1] 问题. 该技术可以与行人检测 [5] 、目标跟踪 [2] 等技术结合应用于智能视频监控、智能安保、智能交通等领域. 在现实场景中, 如刑侦破案以及特定场景下的识人、寻人问题上具有广泛的应用前景.…”

unclassified