Improving person re-identification by attribute and identity learning

Abstract

Person re-identification (re-ID) and attribute recognition share a common target at the pedestrian description. Their difference consists in the granularity. Attribute recognition focuses on local aspects of a person while person re-ID usually extracts global representations. Considering their similarity and difference, this paper proposes a very simple convolutional neural network (CNN) that learns a re-ID embedding and predicts the pedestrian attributes simultaneously. This multi-task method integrates an ID classification loss and a number of attribute classification losses, and back-propagates the weighted sum of the individual losses.Albeit simple, we demonstrate on two pedestrian benchmarks that by learning a more discriminative representation, our method significantly improves the re-ID baseline and is scalable on large galleries. We report competitive re-ID performance compared with the state-of-the-art methods on the two datasets.

AANet: Attribute Attention Network for Person Re-Identifications

“…408 distractor IDs are also included in the dataset. There are a total of 23 attributes annotated by Lin et al [17]. We use all attributes, but with modification to the clothing color attributes.…”

“…The accuracy of attribute classification of the proposed AANet is compared with APR [17] in Table 4. APR [17] is provided by Lin et al, the author who annotated the DukeMTMC-reID and Market1501 datasets with person attributes.…”

“…The accuracy of attribute classification of the proposed AANet is compared with APR [17] in Table 4. APR [17] is provided by Lin et al, the author who annotated the DukeMTMC-reID and Market1501 datasets with person attributes. Since AANet employs localized attribute features to enhance network learning, we obtained better representations and outperforms APR in every attribute prediction.…”

Unsupervised Person Re-identification via Cross-Camera Similarity Exploration

“…Most re-ID methods are in a supervised manner, in which sufficient labeled images are given. Recently, with the developing of deep learning approaches [36,35,34], methods with convolutional neural networks have dominated the re-ID community [12,26,45,46,25,16]. Specifically, methods proposed to learn discriminative features from parts of pedestrian images achieve impressive performance [24,8,23].…”

An Empirical Study of Person Re-Identification with Attributes

“…Visual attributes such as clothing, hair-style, shoe-type, accessories etc, have been used as mid-level feature representation for re-ID in a number of works [16], [30]- [33]. In [31], the authors propose an attribute-recognition network which combines ID classification and attribute classification losses. Because the attributes are human-interpretable, they are useful for zero-shot re-ID [16], [17].…”