期刊
JOURNAL OF VISUAL COMMUNICATION AND IMAGE REPRESENTATION
卷 60, 期 -, 页码 51-58出版社
ACADEMIC PRESS INC ELSEVIER SCIENCE
DOI: 10.1016/j.jvcir.2019.01.010
关键词
Person re-identification; Classification; Feature embedding; CNN; Hypersphere
资金
- Public Projects of Zhejiang Province, China [LGF18F030002]
- National Natural Science Foundation of China [61633019]
Many current successful Person Re-Identification (ReID) methods train a model with the softmax loss function to classify images of different persons and obtain the feature vectors at the same time. However, the underlying feature embedding space is ignored. In this paper, we use a modified softmax function, termed Sphere Softmax, to solve the classification problem and learn a hypersphere manifold embedding simultaneously. A balanced sampling strategy is also introduced. Finally, we propose a convolutional neural network called SphereRelD adopting Sphere Softmax and training a single model endto-end with a new warming-up learning rate schedule on four challenging datasets including Market-1501, DukeMTMC-relD, CHHK-03, and CUHK-SYSU. Experimental results demonstrate that this single model outperforms the state-of-the-art methods on all four datasets without fine-tuning or re-ranking. For example, it achieves 94.4% rank-1 accuracy on Market-1501 and 83.9% rank-1 accuracy on DukeMTMC-relD. The code and trained weights of our model will be released. (C) 2019 Published by Elsevier Inc.
作者
我是这篇论文的作者
点击您的名字以认领此论文并将其添加到您的个人资料中。
推荐
暂无数据