UCT: Learning Unified Convolutional Networks for Real-Time Visual Tracking
UCT: Learning Unified Convolutional Networks for Real-Time Visual Tracking
复制标题
DOI：
10.1109/iccvw.2017.231
复制
发表时间：
2017-11
期刊：
影响因子：
--
通讯作者：
Zheng Zhu;Guan Huang;Wei Zou;Dalong Du;Chang Huang
中科院分区：
文献类型：
--
作者：
Zheng Zhu;Guan Huang;Wei Zou;Dalong Du;Chang Huang
基于卷积神经网络 (CNN) 的跟踪方法在最近的基准测试中表现出了良好的性能。尽管如此，所选择的 CNN 特征总是在不同的任务中进行预训练，并且跟踪系统中的各个组件是单独学习的，因此所实现的跟踪性能可能不是最优的。此外，大多数跟踪器并不是针对实时应用程序而设计的，因为它们的特征提取非常耗时且优化细节也很复杂。在本文中，我们提出了一种端到端框架来学习卷积特征并同时执行跟踪过程，即统一卷积跟踪器（UCT）。具体来说，UCT 将特征提取器和跟踪过程（岭回归）都视为卷积运算并联合训练它们，使学习的 CNN 特征与跟踪过程紧密耦合。在在线跟踪中，通过引入峰值噪声比（PNR）准则，提出了一种有效的更新方法，并通过将尺度分支合并到网络中来有效地处理尺度变化。所提出的方法可实现卓越的跟踪性能，同时保持实时速度。标准 UCT 和 UCT-Lite 可以分别以 41 FPS 和 154 FPS 跟踪通用对象，无需进一步优化。在四个具有挑战性的基准跟踪数据集上进行了实验：OTB2013、OTB2015、VOT2014 和 VOT2015，与其他实时跟踪器相比，我们的方法在这些基准上取得了最先进的结果。
Convolutional neural networks (CNN) based tracking approaches have shown favorable performance in recent benchmarks. Nonetheless, the chosen CNN features are always pre-trained in different task and individual components in tracking systems are learned separately, thus the achieved tracking performance may be suboptimal. Besides, most of these trackers are not designed towards realtime applications because of their time-consuming feature extraction and complex optimization details. In this paper, we propose an end-to-end framework to learn the convolutional features and perform the tracking process simultaneously, namely, a unified convolutional tracker (UCT). Specifically, The UCT treats feature extractor and tracking process (ridge regression) both as convolution operation and trains them jointly, enabling learned CNN features are tightly coupled to tracking process. In online tracking, an efficient updating method is proposed by introducing peak-versus-noise ratio (PNR) criterion, and scale changes are handled efficiently by incorporating a scale branch into network. The proposed approach results in superior tracking performance, while maintaining real-time speed. The standard UCT and UCT-Lite can track generic objects at 41 FPS and 154 FPS without further optimization, respectively. Experiments are performed on four challenging benchmark tracking datasets: OTB2013, OTB2015, VOT2014 and VOT2015, and our method achieves state-of-the-art results on these benchmarks compared with other real-time trackers.