Princeton Tracking Benchmark

Evaluation Result

You can see the detailed evaluation (in .json file) by clicking the title of algorithm. In the ".json" file file, total, type I,II,III are error rate, the successful rate are 1- error rate.

Algorithm Avg.
Rank
target type target size movement oclussion motion type
humananimalrigidlargesmallslowfastyesnopassiveactive
3D-T[1]381.4(1)64.2(5)73.3(7)79.9(1)71.2(2)75.1(6)74.9(1)72.5(1)78.3(5)79.0(3)73.5(1)
ASKCF-OHT[2]9.0951.7(8)49.8(12)72.2(9)58.7(9)59.0(8)67.1(9)55.6(8)52.1(9)68.1(11)71.9(9)54.0(8)
CT[3]21.3631.1(21)46.7(18)36.9(24)39.0(19)34.4(23)48.6(21)31.5(23)23.3(25)54.3(18)42.1(22)34.2(21)
DS-KCF[4]5.0967.0(4)61.2(7)76.4(3)68.8(6)69.7(5)75.4(5)66.9(5)63.3(5)77.6(6)78.8(4)65.7(6)
DSKCF-CPP[5]5.0964.5(5)64.3(4)74.3(5)66.3(7)69.4(6)76.5(2)64.7(7)60.1(7)79.0(4)79.6(2)63.7(7)
DSKCF_SHAPE[6]3.5570.9(3)70.8(2)73.6(6)73.9(3)70.3(3)76.2(4)70.1(3)64.9(3)81.4(2)77.4(6)69.8(4)
Dhog[7f]15.5543.3(14)48.3(15)55.9(14)47.2(16)50.3(15)52.7(18)47.5(14)38.4(16)63.5(15)54.3(19)46.9(15)
ExampleResult2017-03-051488720222[8]20.6432.9(19)36.3(22)45.2(20)33.8(22)41.4(20)45.8(23)35.1(20)34.1(18)43.7(22)47.4(21)34.6(20)
KCF[9]12.8241.8(15)50.4(11)64.9(11)48.4(14)54.7(12)65.0(11)46.9(15)40.6(14)67.7(12)64.5(12)47.3(14)
LDPSTRUCK[10]1346.2(12)59.1(8)54.5(17)51.6(13)52.0(14)56.2(17)50.1(12)39.8(15)68.4(9)56.4(15)50.1(11)
LDP_SVT[11]12.0946.8(11)58.8(9)55.8(15)52.9(11)52.3(13)56.3(16)51.1(11)41.0(13)68.4(10)58.2(14)50.4(10)
MIL[12]22.1832.2(20)37.2(21)38.3(23)36.6(21)34.6(22)45.5(24)31.5(22)25.6(23)49.0(20)40.4(25)33.6(23)
OAPF[13]2.8264.2(6)84.8(1)77.2(2)72.7(4)73.4(1)85.1(1)68.4(4)64.4(4)85.1(1)77.7(5)71.4(2)
OF[7g]2617.9(26)11.4(26)23.4(26)20.1(26)17.5(26)18.1(26)18.8(26)15.9(26)22.3(26)23.4(26)16.8(26)
PCdet[7i]14.7340.6(16)42.1(19)61.7(13)55.4(10)43.6(19)58.5(12)44.8(16)46.3(11)52.0(19)64.9(11)42.6(16)
PCdet_flow[7h]8.9150.5(9)51.6(10)72.7(8)63.4(8)55.5(10)73.9(7)53.0(9)55.0(8)64.4(13)75.5(7)52.7(9)
PCflow[7j]21.4535.2(18)29.1(25)43.6(22)42.2(18)33.2(24)47.2(22)33.1(21)32.4(20)43.5(23)41.3(24)35.5(19)
RGB[7e]19.6426.7(24)40.9(20)54.7(16)31.9(24)46.0(17)50.5(20)35.7(19)34.8(17)46.8(21)56.2(16)33.7(22)
RGBD+OF[7c]5.6463.9(7)65.3(3)74.5(4)71.5(5)65.5(7)73.4(8)65.9(6)60.1(6)79.0(3)74.0(8)65.8(5)
RGBDOcc+OF[7a]374.0(2)62.6(6)78.4(1)78.1(2)69.7(4)76.3(3)72.2(2)72.0(2)75.2(7)82.3(1)70.0(3)
RGBOF[7d]12.6447.1(10)47.0(16)63.6(12)47.4(15)57.5(9)56.7(15)51.8(10)46.9(10)61.9(17)63.4(13)49.3(12)
SAMF+Depth[14]11.3644.8(13)49.6(13)67.0(10)52.4(12)55.2(11)65.2(10)49.5(13)41.1(12)71.6(8)66.3(10)49.3(13)
SemiB[15]24.6422.5(25)33.0(24)32.7(25)24.0(25)31.6(25)38.2(25)24.4(25)25.1(24)32.7(25)41.9(23)23.2(25)
Struck[16]17.0935.4(17)47.0(17)53.4(19)45.0(17)43.9(18)58.0(13)39.0(17)30.4(21)63.5(14)54.4(18)40.6(17)
TLD[17]21.9129.0(23)35.1(23)44.4(21)32.5(23)38.5(21)51.6(19)29.7(24)33.8(19)38.7(24)50.2(20)30.5(24)
VTD[18]17.7330.9(22)48.8(14)53.9(18)38.6(20)46.2(16)57.3(14)37.2(18)28.3(22)63.1(16)54.9(17)38.5(18)

References

[1]Adel Bibi, Tianzhu Zhang, and Bernard Ghanem,  3D Part-Based Sparse Tracker with Automatic Synchronization and Registration  Conference on Computer Vision and Pattern Recognition (CVPR2016)
[2]J.T. Pi and K.L. Hu,    
[3]K. Zhang, L. Zhang, and M.-H. Yang,  Real-time compressive tracking.  ECCV, 2012
[4]M. Camplani, S. Hannuna, M. Mirmehdi, D. Damen, A. Paiement, L. Tao, T. Burghardt,  Real-time RGB-D Tracking with Depth Scaling Kernelised Correlation Filters and Occlusion Handling.  BMVC 2015
[5]S. Hannuna, M. Camplani, J. Hall, M. Mirmehdi, D. Damen, T. Burghardt, A. Paiement, L. Tao,  DS-KCF: A real-time tracker for RGB-D data  Journal of Real-Time Image Processing
[6]S. Hannuna, M. Camplani, J. Hall, M. Mirmehdi, D. Damen, T. Burghardt, A. Paiement, L. Tao,  DS-KCF: A real-time tracker for RGB-D data  Journal of Real-Time Image Processing
[7]S. Song and J. Xiao,  Tracking Revisited using RGBD Camera: Baseline and Benchmark
7a. RGBD HoG detection+ optical flow + Occlusion handle.
7c. RGBD HoG detection+ optical flow.
7d. RGBD HoG detection.
7e. RGB HoG detection.
7f. D HoG detection.
7g. Optical flow.
7h. Point Cloud (detection+ flow).
7i. Point Cloud detection.
7j. Point Cloud flow.
[8]Anonymous,    
[9]Test run by Massimo Camplani, authors are: J. F. Henriques, R. Caseiro, P. Martins, J. Batista,  High-Speed Tracking with Kernelized Correlation Filters  TPAMI 2015
[10]Awwad Sari, Fairouz Hussein, and Massimo Piccardi,  Local Depth Patterns for Tracking in Depth Videos  Proceedings of the 23rd Annual ACM Conference on Multimedia Conference. ACM, 2015
[11]Sari Awwad and Massimo Piccardi,  Prototype-based budget maintenance for tracking in depth videos  Multimedia Tools and Applications Journal (2016)
[12]B. Babenko, M.-H. Yang, and S. Belongie,  Visual Tracking with Online Multiple Instance Learning  CVPR, 2009
[13]K. Meshgi, S. Maeda, S. Oba, H. Skibbe, Y. Li, and S. Ishii,  Occlusion aware particle filter tracker to handle complex and persistent occlusions  CVIU, 2015
[14]Yang Li and Jianke Zhu,  A Scale Adaptive Kernel Correlation Filter Tracker with Feature Integration  ECCV 2014 Workshop
[15]H. Grabner, C. Leistner, and H. Bischof,  Semi-supervised on-line boosting for robust tracking  ECCV, 2008
[16]S. Hare, A. Saffari, and P. H. S. Torr.,  Struck: Structured output tracking with kernels.  In ICCV, 2011.
[17]Z. Kalal, K. Mikolajczyk, and J. Matas,  Tracking-learning detection  PAMI, 2012.
[18]J. Kwon and K. M. Lee,  Visual tracking decomposition  CVPR, 2010.