SCAN: Spatial and Channel Attention Network for Vehicle Re-Identification

Proceedings of the 28th ACM International Conference on Multimedia

et al. 2020

Vehicle re-identification aims to identify the same vehicle across different surveillance cameras and plays an important role in public security. Existing approaches mainly focus on exploring informative regions or learning an appropriate distance metric. However, they not only neglect the inherent structured relationship between discriminative regions within an image, but also ignore the extrinsic structured relationship among images. The inherent and extrinsic structured relationships are crucial to learning effective vehicle representation. In this paper, we propose a Structured Graph ATtention network (SGAT) to fully exploit these relationships and allow the message propagation to update the features of graph nodes. SGAT creates two graphs for one probe image. One is an inherent structured graph based on the geometric relationship between the landmarks that can use features of their neighbors to enhance themselves. The other is an extrinsic structured graph guided by the attribute similarity to update image representations. Experimental results on two public vehicle re-identification datasets including VeRi-776 and VehicleID have shown that our proposed method achieves significant improvements over the state-of-the-art methods. CCS CONCEPTS • Information systems → Top-k retrieval in databases; • Computing methodologies → Visual content-based indexing and retrieval.

Section: Experiments Resultsmentioning

confidence: 99%

Section: Experiments Resultsmentioning

confidence: 99%

A Structured Graph Attention Network for Vehicle Re-Identification

Zhu

Zha

Proceedings of the 28th ACM International Conference on Multimedia

et al. 2020

“…We implemented our framework in Ten-sorflow1.0 on a computer with an i7 6700 CPU, an NVIDIA Titan X GPU and 32 GB of RAM. For the re-identification networks [32] and social LSTM network [1], we use the implementations released by the authors. Evaluation.…”

Section: Methodsmentioning

confidence: 99%

“…For each target in frame − 1, the weights of the targets are generated by calculating the appearance similarity between the target and others targets around. The appearance feature of target x is generated by a object re-identification networks [32]. With the weights, the appearance similarity between targets in frame − 1 and the objects in frame would be generated.…”

Section: The Motion Consistencymentioning

confidence: 99%

Self-balance Motion and Appearance Model for Multi-object Tracking in UAV

Proceedings of the ACM Multimedia Asia

et al. 2019

Self Cite

Under the tracking-by-detection framework, multi-object tracking methods try to connect object detections with target trajectories by reasonable policy. Most methods represent objects by the appearance and motion. The inference of the association is mostly judged by a fusion of appearance similarity and motion consistency. However, the fusion ratio between appearance and motion are often determined by subjective setting. In this paper, we propose a novel self-balance method fusing appearance similarity and motion consistency. Extensive experimental results on public benchmarks demonstrate the effectiveness of the proposed method with comparisons to several state-of-the-art trackers.

“…Supervised person ReID task has been widely studied and many methods have been proposed from many respects, such as parts feature extraction [20,28,34,41,52,53], distance metric learning [3,12,29,48,63], attention learning [2,35], etc. However, manual person ID annotation is expensive and rarely available in real-world application.…”

Section: Supervised Person Re-identificationmentioning

confidence: 99%

Domain Adaptive Person Re-Identification via Coupling Optimization

Liu

Proceedings of the 28th ACM International Conference on Multimedia

2020

Self Cite

Domain adaptive person Re-Identification (ReID) is challenging owing to the domain gap and shortage of annotations on target scenarios. To handle those two challenges, this paper proposes a coupling optimization method including the Domain-Invariant Mapping (DIM) method and the Global-Local distance Optimization (GLO), respectively. Different from previous methods that transfer knowledge in two stages, the DIM achieves a more efficient one-stage knowledge transfer by mapping images in labeled and unlabeled datasets to a shared feature space. GLO is designed to train the ReID model with unsupervised setting on the target domain. Instead of relying on existing optimization strategies designed for supervised training, GLO involves more images in distance optimization, and achieves better robustness to noisy label prediction. GLO also integrates distance optimizations in both the global dataset and local training batch, thus exhibits better training efficiency. Extensive experiments on three large-scale datasets, i.e., Market-1501, DukeMTMC-reID, and MSMT17, show that our coupling optimization outperforms state-of-the-art methods by a large margin. Our method also works well in unsupervised training, and even outperforms several recent domain adaptive methods.