Learning to Associate: HybridBoosted Multi-Target Tracker for Crowded Scene

1. Learning to Associate: HybridBoosted Multi-Target Tracker for Crowded Scene Yuan Li, Chang Huang and Ram Nevatia

2. Yuan Li

3. Outline • introduction • Related work • MAP formulation • Affinity model • Results • Conclusion

6. overview

7. STAGE 1 STAGE 2 STAGE 3 STAGE 4

8. Introduction • learning-based hierarchical approach of multi-target tracking • HybridBoost algorithm-hybrid loss function • association of tracklet is formulated as a joint problem of ranking and classification

9. ranking • the ranking part aims to rank correct tracklet associations higher than other alternatives

10. classification • the classification part is responsible to reject wrong associations when no further association should be done

11. HybridBoost • combines the merits of the RankBoost algorithm and the AdaBoost algorithm .

12. adaboost

13. RankBoost

14. Related work • the earliest works look at a longer period of time in contrast to frame-by-frame tracking. • To overcome this, a category of DataAssociation based Tracking algorithm • there has been no use of machine learning algorithmin building the affinity model.

15. MAP formulation • Robust Object Tracking by Hierarchical Association of Detection Responses • ours

16. MAP formulation v1 • R = {ri} the set of all detection responses

17. MAP formulation v1(cont.) • tracklet association

18. MAP formulation v1(cont.)

19. MAP formulation v2

20. MAP formulation v2(cont.) • Inner cost • Transition cost

21. MAP formulation v2(cont.) • With these ,we can rewrite it

22. Affinity model • Hybridboostalgorithm • Feature pool and weak learner • Training process

23. Hybridboostalgorithm • Ie. T2 T1 T3

24. Hybridboostalgorithm(cont.)

25. Loos function • initial

27. Hybridboostalgorithm

29. Feature pool and weak learner

30. Training process • T:tracklet set from the previous stage • G:groundtruth track set

31. Training process(cont) • For each Ti ∈ T, if • connecting Ti’stail to the head of some other tracklet

32. Training process(cont) • connecting Ti’s head to the tail of some other tracklet before Ti which is also matched to G

33. Ranking sample set

34. Binary sample set

35. Training process(cont.) • use thegroundtruthG and the tracklet set Tk−1 obtained from stagek − 1 to generate ranking and binary classification samples • learn a strong ranking classifier Hkby the HybridBoost algorithm • UsingHk as the affinity model to perform association on Tk−1and generate Tk

36. Experimental results • Implementation details • Evaluation metrics • Analysis of the training process • Tracking performance

37. Implementation details • dual-threshold strategy to generate short but reliable tracklets • four stages of association • maximum allowed frame gap 16, 32, 64 and 128 • a strong ranking classifier H with 100 weak ranking classifiers • Β=0.75 • ζ = 0

38. Evaluation metrics

39. track fragments &ID switches • Traditional ID switch:“two tracks exchanging their ids”. • ID switch : a tracked trajectory changing its matched GT ID • track fragments:more strict

40. compare

41. Best features • Motion smoothness (feature type 13 or 14) • color histogram similarity (feature 4) • number of miss detected frames in the gap between the two trackelts (feature 7 or 9).

42. Strong ranking classifier output

43. Choice of β

44. Tracking performance

45. Conclusion and future work • Use HybridBoost algorithm to learn the affinity model as a joint problem of ranking and classification • The affinity model is integrated in a hierarchical data association framework to track multiple targets in very crowded scenes.

47. Thank you

48. problem • tracklet ?affinity model?圓圈?路徑? • automatically selectamong various features andcorresponding non-parametric models? Rankboost ? Adaboost? 匈牙利演算法