Abstract | ||
---|---|---|
In this paper, we propose a multi-branch model of spatial global attention mechanism to perform the task of pedestrian recognition. The main research direction of the previous pedestrian re-recognition is detection-recognition, which only correlates the appearance information between objects to ascertain the object's target. This paper proposes a multi-branch model result, using the spatial information of the recognition object, proposes a multi-branch network based on the global spatial attention mechanism, and establishes the object connection for pedestrian recognition through the spatial relationship between the main branch and the auxiliary branch. Another branch focuses on the longitudinal association of pedestrians through the attention mechanism, establishes the connection between key points of pedestrians, and optimizes the Mahalanobis distance between key points. The method proposed under this paper has improved accuracy and speed compared with preceding network performance in several pedestrian re-identification databases. (C) 2021 Elsevier Inc. All rights reserved. |
Year | DOI | Venue |
---|---|---|
2022 | 10.1016/j.bdr.2021.100302 | BIG DATA RESEARCH |
Keywords | DocType | Volume |
Multi-branch network, Re-identification, Attention mechanism | Journal | 27 |
ISSN | Citations | PageRank |
2214-5796 | 0 | 0.34 |
References | Authors | |
0 | 4 |
Name | Order | Citations | PageRank |
---|---|---|---|
Feng Hong | 1 | 0 | 0.34 |
Chang-Hua Lu | 2 | 0 | 0.34 |
Tao Wang | 3 | 0 | 0.68 |
Weiwei Jiang | 4 | 0 | 0.34 |