期刊
JOURNAL OF REAL-TIME IMAGE PROCESSING
卷 17, 期 5, 页码 1647-1657出版社
SPRINGER HEIDELBERG
DOI: 10.1007/s11554-019-00922-6
关键词
Computer vision; Object tracking; Siamese network
类别
资金
- Natural Science Foundation of Jiangsu Province [BK20151102]
- Natural Science Foundation of China [61802058]
- Ministry of Education Key Laboratory of Machine Perception, Peking University [K-2016-03]
- Open Project Program of the Ministry of Education Key Laboratory of Underwater Acoustic Signal Processing, Southeast University [UASP1502]
Considering that most deep learning based trackers capture accurate locations for targets at the expense of consuming much time in training phrase, in this paper we present a new powerful tracker using the Siamese network which can be implemented with low computation resource. Our proposed tracker can track targets accurately by a fine-tuned model which is convenient to train. During the tracking, we apply a new sampling method that is independent of training called action-selection to conduct selective and flexible sampling step by step with a variable stride, by which we can get bounding boxes with varied aspect radio. By verifying its performance on online tracking benchmarks, it turns out that our tracker achieves higher accuracy than most traditional trackers. In addition, our tracker operates at frame-rates beyond real-time.
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