Global-Local Dynamic Feature Alignment Network for Person Re-Identification

Abstract

We propose a new method name LSA that achieves dynamic alignment of local features by setting sliding windows for their local stripes• LSA can effectively suppress spatially misalignment and the noises from unshared regions.• LSA does not require the additional auxiliary pose information.• We design a Global-Local Dynamic Feature Alignment Network (GLDFA-Net) framework, which contains two branches, global and local.• LSA is introduced into the local branch of GLDFA-Net to guide the computation of the distance metric.

Deep learning-based person re-identification methods: A survey and outlook of recent works

“…Li et al [57] used balanced attention Fig. 7: Schematic of Global-Local Dynamic Feature Alignment Network(GLDFA-Net) architecture [84] convolutional neural networks to maximize the complementary information of attention features at different scales to solve the person Re-ID problem for arbitrary unaligned images. Chen et al [6] proposed a higher-order attention module,which models the complex higher-order information in the attention mechanism to mine discriminative attention features among pedestrians.…”

“…Wang et al [110] proposed a multi-granularity feature learning strategy with global and local information, including one branch for global feature learning and two branches for local feature learning. Ming et al [84] design a global-local dynamic feature alignment network (GLDFA-Net) framework, which contains both global and local branches, their network structure is shown in Figure 7. Local sliding alignment(LSA) strategy is introduce into the local branch of GLDFA-Net to guide the computation of distance metrics, which can further improve the accuracy of the testing phase.…”