DC-Former: Diverse and Compact Transformer for Person Re-Identification

Abstract

In person re-identification (re-ID) task, it is still challenging to learn discriminative representation by deep learning, due to limited data. Generally speaking, the model will get better performance when increasing the amount of data. The addition of similar classes strengthens the ability of the classifier to identify similar identities, thereby improving the discrimination of representation. In this paper, we propose a Diverse and Compact Transformer (DC-Former) that can achieve a similar effect by splitting embedding space into multiple diverse and compact subspaces. Compact embedding subspace helps model learn more robust and discriminative embedding to identify similar classes. And the fusion of these diverse embeddings containing more fine-grained information can further improve the effect of re-ID. Specifically, multiple class tokens are used in vision transformer to represent multiple embedding spaces. Then, a self-diverse constraint (SDC) is applied to these spaces to push them away from each other, which makes each embedding space diverse and compact. Further, a dynamic weight controller (DWC) is further designed for balancing the relative importance among them during training. The experimental results of our method are promising, which surpass previous state-of-the-art methods on several commonly used person re-ID benchmarks. https://github.com/ant-research/Diverseand-Compact-Transformer

A method based on difference guide and feature self-enhancement for clothes-changing person re-identification

“…[27] develope a model called Knowledge Refresh and Consolidation (KRC) which enhanced by the introduction of a dynamic memory model and an adaptive working model that enables bidirectional transfer of knowledge and a knowledge consolidation scheme that operates on a dual space. [28] propose a Diverse and Compact Transformer (DC-Former) that can split embedding space into multiple diverse and compact subspaces, and self-diverse constraint (SDC) are imposed on subspaces through multiple class labelling, thus making each embedding space diverse and compact. [29] establish a new benchmark framework called TransReID, in which the transformer framework is employed.…”

An Orientation-Aware Attention Network for Person Re-Identification

“…We compare the performance of MPGA-Net with recent state-of-the-art person ReID methods on CUHK03-NP [31,32], Market1501 [33], DukeMTMC-reID [34], and MSMT17 [35] in Table 1, including methods based on ResNet [7,14,22,23,26,[37][38][39][40][41][42][43][44], self-constructed network [6], neural architecture search [45], and tranformer [20,[46][47][48]. Overall, our proposed MPGA-Net outperforms the state-of-the-art networks or achieves comparable performance.…”

TSPT ReID: Triple Streams with Pos-insertion Transformer for Person Re-identification

“…Zhu et al [35] proposes embedding a learnable cls_token (CLS) with a single-head auto-alignment structure for accurate human body part representation. DC-Former [36] incorporates a Self-diverse Constraint Layer after attention, enhancing feature vector distances to bolster loss convergence. Moreover, PAT [37] employes self-attention and cross-attention for feature encoding and decoding, facilitating global and local human body feature extraction.…”