Discriminative Feature Representation for Person Re-identification by Batch-contrastive Loss

Guopeng Zhang, Jinhua Xu
Proceedings of The 10th Asian Conference on Machine Learning, PMLR 95:208-219, 2018.

Abstract

In the past few years, person re-identification (reID) has developed rapidly due to the success of deep convolutional neural networks. The softmax loss function is an important component for learning discriminative features. However, the classifier trained by the softmax loss is difficult to distinguish the hard samples. In this work, we introduce a new auxiliary loss function, called batch-contrastive loss, for person reID to further separate the features of different identities and pulls the features of same identity closer. Furthermore, the proposed loss function does not rely on the pairwise or triplet sampling which is commonly used in the Siamese model. We test our loss function on two large-scale person reID benchmarks, Market-1501 and DukeMTMC datasets. Under the combination of the batch-contrastive loss and the softmax loss, even only employing the generic L2-distance metric, we can achieve competitive results among the state-of-the-arts.

Cite this Paper

BibTeX
@InProceedings{pmlr-v95-zhang18b, title = {Discriminative Feature Representation for Person Re-identification by Batch-contrastive Loss}, author = {Zhang, Guopeng and Xu, Jinhua}, booktitle = {Proceedings of The 10th Asian Conference on Machine Learning}, pages = {208--219}, year = {2018}, editor = {Zhu, Jun and Takeuchi, Ichiro}, volume = {95}, series = {Proceedings of Machine Learning Research}, month = {14--16 Nov}, publisher = {PMLR}, pdf = {http://proceedings.mlr.press/v95/zhang18b/zhang18b.pdf}, url = {https://proceedings.mlr.press/v95/zhang18b.html}, abstract = {In the past few years, person re-identification (reID) has developed rapidly due to the success of deep convolutional neural networks. The softmax loss function is an important component for learning discriminative features. However, the classifier trained by the softmax loss is difficult to distinguish the hard samples. In this work, we introduce a new auxiliary loss function, called batch-contrastive loss, for person reID to further separate the features of different identities and pulls the features of same identity closer. Furthermore, the proposed loss function does not rely on the pairwise or triplet sampling which is commonly used in the Siamese model. We test our loss function on two large-scale person reID benchmarks, Market-1501 and DukeMTMC datasets. Under the combination of the batch-contrastive loss and the softmax loss, even only employing the generic L2-distance metric, we can achieve competitive results among the state-of-the-arts.} }
Endnote
%0 Conference Paper %T Discriminative Feature Representation for Person Re-identification by Batch-contrastive Loss %A Guopeng Zhang %A Jinhua Xu %B Proceedings of The 10th Asian Conference on Machine Learning %C Proceedings of Machine Learning Research %D 2018 %E Jun Zhu %E Ichiro Takeuchi %F pmlr-v95-zhang18b %I PMLR %P 208--219 %U https://proceedings.mlr.press/v95/zhang18b.html %V 95 %X In the past few years, person re-identification (reID) has developed rapidly due to the success of deep convolutional neural networks. The softmax loss function is an important component for learning discriminative features. However, the classifier trained by the softmax loss is difficult to distinguish the hard samples. In this work, we introduce a new auxiliary loss function, called batch-contrastive loss, for person reID to further separate the features of different identities and pulls the features of same identity closer. Furthermore, the proposed loss function does not rely on the pairwise or triplet sampling which is commonly used in the Siamese model. We test our loss function on two large-scale person reID benchmarks, Market-1501 and DukeMTMC datasets. Under the combination of the batch-contrastive loss and the softmax loss, even only employing the generic L2-distance metric, we can achieve competitive results among the state-of-the-arts.
APA
Zhang, G. & Xu, J.. (2018). Discriminative Feature Representation for Person Re-identification by Batch-contrastive Loss. Proceedings of The 10th Asian Conference on Machine Learning, in Proceedings of Machine Learning Research 95:208-219 Available from https://proceedings.mlr.press/v95/zhang18b.html.

Related Material