AI helps you reading Science
AI generates interpretation videos
AI extracts and analyses the key points of the paper to generate videos automatically
AI parses the academic lineage of this thesis
AI extracts a summary of this paper
Correlation Tracking Via Joint Discrimination And Reliability Learning
2018 IEEE/CVF CONFERENCE ON COMPUTER VISION AND PATTERN RECOGNITION (CVPR), (2018): 489-497
- Visual tracking is a hot topic for its wide applications in many computer vision tasks, such as video surveillance, behaviour analysis, augmented reality, to name a few.
- By multiplying the filter with a binary mask, the tracker is able to generate the real training samples having the same size as the target object, and better suppressing the background regions
- This method has two limitations: first, it exploits the augmented Lagrangian method for model learning, which limits the model extension; second, even though the background region outside the bounding box is suppressed, the tracker may be influenced by the background region inside the bounding box
- Visual tracking is a hot topic for its wide applications in many computer vision tasks, such as video surveillance, behaviour analysis, augmented reality, to name a few
- As the correlation filter takes the entire image as the positive sample and the cyclically shifted images as negative ones, the learned filters are likely to be influenced by the background regions
- We introduce a local response consistency constraint for the base filter, which constrains that each sub-region of the target has similar importance
- The reliability information is separated from the base filter
- We consider the reliability information in the filter, which is jointly learned with the base filter
- Compared to the existing correlation filter-based methods, our tracker is insusceptible to the non-uniform distributions of the feature map, and can better suppress the background regions
- Let y=[y1, y2, ..., yK ]⊤ ∈ RK×1 denote gaussian shaped response, and xd ∈ RK×1 be the input vector for the d-th channel, the correlation filter learns the optimal w by optimizing the following formula: K.
- K=1 d=1 where xk,d is the k-step circular shift of the input vector xd, yk is the k-th element of y, w= w1⊤, w2⊤, ..., wD⊤ ⊤ where wd ∈ RK×1 stands for the filter of the d-th channel.
- The authors demonstrate the effectiveness of the proposed tracker on the OTB-2013 , OTB-2015  and VOT-2016  benchmark datasets.
- Since the method jointly considers both discrimination and reliability for tracking, the authors denote it as DRT for clarity.
- The proposed DRT method is mainly implemented in MATLAB and is partially accelerated with the Caffe toolkit .
- The authors use a relatively small learning rate ω (i.e. 0.011) for first 10 frames to avoid model degradation with limited training samples, and use a larger one (i.e. 0.02) in the following tracking process.
- One implementation of the tracker can be found in https://github.com/cswaynecool/DRT
- The authors clearly consider the discrimination and reliability information in the correlation filter (CF) formula and rewrite the filter weight as the element-wise product of a base filter and a reliability weight map.
- The authors introduce a local response consistency constraint for the base filter, which constrains that each sub-region of the target has similar importance.
- By this means, the reliability information is separated from the base filter.
- Compared to the existing CF-based methods, the tracker is insusceptible to the non-uniform distributions of the feature map, and can better suppress the background regions.
- Extensive experiments demonstrate that the proposed tracker outperforms the state-of-the-art algorithms over all three benchmarks
- Table1: Performance evaluation of different state-of-the-art trackers in the VOT-2016 dataset. In this dataset, we compare our DRT method with the top 10 trackers. The best two results are marked in red and blue bold fonts, respectively
- This paper is partially supported by the Natural Science Foundation of China #61502070, #61725202, #61472060
- Chong Sun and Ming-Hsuan Yang are also supported in part by NSF CAREER (No 1149783), gifts from Adobe, Toyota, Panasonic, Samsung, NEC, Verisk, and Nvidia
- L. Bertinetto, J. Valmadre, S. Golodetz, O. Miksik, and P. H. Torr. Staple: Complementary learners for real-time tracking. In CVPR, 2016.
- L. Bertinetto, J. Valmadre, J. F. Henriques, A. Vedaldi, and P. H. Torr. Fully-convolutional siamese networks for object tracking. In ECCV, 2016.
- D. S. Bolme, J. R. Beveridge, B. A. Draper, and Y. M. Lui. Visual object tracking using adaptive correlation filters. In CVPR, 2010.
- K. Chatfield, K. Simonyan, A. Vedaldi, and A. Zisserman. Return of the devil in the details: Delving deep into convolutional nets. In BMVC, 2014.
- M. Danelljan, G. Bhat, F. S. Khan, and M. Felsberg. Eco: Efficient convolution operators for tracking. In CVPR, 2017.
- M. Danelljan, G. Hager, F. Khan, and M. Felsberg. Accurate scale estimation for robust visual tracking. In BMVC, 2014.
- M. Danelljan, G. Hager, F. Shahbaz Khan, and M. Felsberg. Convolutional features for correlation filter based visual tracking. In ICCV Workshops, 2015.
- M. Danelljan, G. Hager, F. Shahbaz Khan, and M. Felsberg. Learning spatially regularized correlation filters for visual tracking. In ICCV, 2015.
- M. Danelljan, G. Hager, F. Shahbaz Khan, and M. Felsberg. Adaptive decontamination of the training set: A unified formulation for discriminative visual tracking. In CVPR, 2016.
- M. Danelljan, A. Robinson, F. S. Khan, and M. Felsberg. Beyond correlation filters: Learning continuous convolution operators for visual tracking. In ECCV, 2016.
- J. F. Henriques, R. Caseiro, P. Martins, and J. Batista. Exploiting the circulant structure of tracking-by-detection with kernels. In ECCV, 2012.
- J. F. Henriques, R. Caseiro, P. Martins, and J. Batista. Highspeed tracking with kernelized correlation filters. IEEE Transactions on Pattern Analysis and Machine Intelligence, 37(3):583–596, 2015.
- Y. Jia, E. Shelhamer, J. Donahue, S. Karayev, J. Long, R. Girshick, S. Guadarrama, and T. Darrell. Caffe: Convolutional architecture for fast feature embedding. In ICMM, 2014.
- H. Kiani Galoogahi, A. Fagg, and S. Lucey. Learning background-aware correlation filters for visual tracking. In CVPR, 2017.
- M. Kristan, J. Matas, A. Leonardis, M. Felsberg, L. Cehovin, G. Fernandez, T. Vojir, G. Hager, G. Nebehay, and R. Pflugfelder. The visual object tracking vot2016 challenge results. In ECCV Workshops, 2016.
- P. Li, D. Wang, L. Wang, and H. Lu. Deep visual tracking: Review and experimental comparison. Pattern Recognition, 76:323–338, 2018.
- Y. Li, J. Zhu, and S. C. Hoi. Reliable patch trackers: Robust visual tracking by exploiting reliable patches. In CVPR, 2015.
- S. Liu, T. Zhang, X. Cao, and C. Xu. Structural correlation filter for robust visual tracking. In CVPR, 2016.
- T. Liu, G. Wang, and Q. Yang. Real-time part-based visual tracking via adaptive correlation filters. In CVPR, 2015.
- A. Lukezic, T. Vojır, L. Cehovin, J. Matas, and M. Kristan. Discriminative correlation filter with channel and spatial reliability. In CVPR, 2017.
- C. Ma, J.-B. Huang, X. Yang, and M.-H. Yang. Hierarchical convolutional features for visual tracking. In ICCV, 2015.
- C. Ma, X. Yang, C. Zhang, and M.-H. Yang. Long-term correlation tracking. In CVPR, 2015.
- H. Nam, M. Baek, and B. Han. Modeling and propagating cnns in a tree structure for visual tracking. arXiv preprint arXiv:1608.07242, 2016.
- J. Nocedal and S. J. Wright. Numerical Optimization. Springer, New York, USA, 2006.
- Y. Qi, L. Qin, J. Zhang, S. Zhang, Q. Huang, and M.-H. Yang. Structure-aware local sparse coding for visual tracking. IEEE Transactions on Image Processing, PP(99):1–1, 2018.
- Y. Qi, S. Zhang, L. Qin, H. Yao, Q. Huang, J. Lim, and M.-H. Yang. Hedged deep tracking. In CVPR, 2016.
- R. Tao, E. Gavves, and A. W. Smeulders. Siamese instance search for tracking. In CVPR, 2016.
- L. Wang, W. Ouyang, X. Wang, and H. Lu. Visual tracking with fully convolutional networks. In ICCV, 2015.
- L. Wang, W. Ouyang, X. Wang, and H. Lu. Stct: Sequentially training convolutional networks for visual tracking. In CVPR, 2016.
- Y. Wu, J. Lim, and M.-H. Yang. Online object tracking: A benchmark. In CVPR, 2013.
- Y. Wu, J. Lim, and M.-H. Yang. Object tracking benchmark. IEEE Transactions on Pattern Analysis and Machine Intelligence, 37(9):1834–1848, 2015.
- J. Zhang, S. Ma, and S. Sclaroff. Meem: robust tracking via multiple experts using entropy minimization. In ECCV, 2014.
- G. Zhu, F. Porikli, and H. Li. Beyond local search: Tracking objects everywhere with instance-specific proposals. In CVPR, 2016.