sa ›› 2018, Vol. 23 ›› Issue (3): 360-.doi: 10.1007/s12204-018-1951-8

• • 上一篇    下一篇

Branch-Activated Multi-Domain Convolutional Neural Network for Visual Tracking

CHEN Yimin (陈一民), LU Rongrong (陆蓉蓉), ZOU Yibo (邹一波), ZHANG Yanhui (张燕辉)   

  1. (School of Computer Engineering and Science, Shanghai University, Shanghai 200444, China)
  • 出版日期:2018-05-31 发布日期:2018-06-17
  • 通讯作者: CHEN Yimin (陈一民) E-mail:ymchen@mail.shu.edu.cn

Branch-Activated Multi-Domain Convolutional Neural Network for Visual Tracking

CHEN Yimin (陈一民), LU Rongrong (陆蓉蓉), ZOU Yibo (邹一波), ZHANG Yanhui (张燕辉)   

  1. (School of Computer Engineering and Science, Shanghai University, Shanghai 200444, China)
  • Online:2018-05-31 Published:2018-06-17
  • Contact: CHEN Yimin (陈一民) E-mail:ymchen@mail.shu.edu.cn

摘要: Convolutional neural networks (CNNs) have been applied in state-of-the-art visual tracking tasks to represent the target. However, most existing algorithms treat visual tracking as an object-speciˉc task. Therefore, the model needs to be retrained for di?erent test video sequences. We propose a branch-activated multi-domain convolutional neural network (BAMDCNN). In contrast to most existing trackers based on CNNs which require frequent online training, BAMDCNN only needs o2ine training and online fine-tuning. Speciˉcally, BAMDCNN exploits category-specific features that are more robust against variations. To allow for learning category-specific information, we introduce a group algorithm and a branch activation method. Experimental results on challenging benchmark show that the proposed algorithm outperforms other state-of-the-art methods. What's more, compared with CNN based trackers, BAMDCNN increases tracking speed.

关键词: convolutional neural network (CNN), category-specific feature, group algorithm, branch activation method

Abstract: Convolutional neural networks (CNNs) have been applied in state-of-the-art visual tracking tasks to represent the target. However, most existing algorithms treat visual tracking as an object-speciˉc task. Therefore, the model needs to be retrained for di?erent test video sequences. We propose a branch-activated multi-domain convolutional neural network (BAMDCNN). In contrast to most existing trackers based on CNNs which require frequent online training, BAMDCNN only needs o2ine training and online fine-tuning. Speciˉcally, BAMDCNN exploits category-specific features that are more robust against variations. To allow for learning category-specific information, we introduce a group algorithm and a branch activation method. Experimental results on challenging benchmark show that the proposed algorithm outperforms other state-of-the-art methods. What's more, compared with CNN based trackers, BAMDCNN increases tracking speed.

Key words: convolutional neural network (CNN), category-specific feature, group algorithm, branch activation method

中图分类号: