1. Learning to Associate: HybridBoosted Multi-Target Tracker for Crowded Scene Present by 陳群元

2. review

4. System Architecture

5. Strong ranking classifier

6. Weak ranking classifier Feature & threshold

7. previous problem The lengths of ground truth tracklets are equal.

8. solution Cut trajectory to tracklet randomly

9. Previous problem The scales of some thresholds are wide. Feature 1Feature 5

10. solution Quantize these features’ threshold with respective bins. Quantize the difference of min and max value with difference bin.

11. Middle-Level Association The middle level association is an iterative process: each round takes the tracklets generated in the previous round as the input and does further association

12. First round input tracklet association –l k is the number of tracklets in S k. corresponding trajectory of S k tracklet association set.

13. MAP problem

14. How to associate Bruce force?

15. Hungarian algorithm

16. Hungarian Algorithm(1) Arrange your information in a matrix with the "people" on the left and the "activity" along the top, with the "cost" for each pair in the middle.

17. Hungarian Algorithm(2) Ensure that the matrix is square by the addition of dummy rows/columns if necessary.

18. Hungarian Algorithm(3) Reduce the rows by subtracting the minimum value of each row from that row.

19. Hungarian Algorithm(4) Reduce the columns by subtracting the minimum value of each column from that column.

20. Hungarian Algorithm(5) Cover the zero elements with the minimum number of lines it is possible to cover them with.

21. Hungarian Algorithm(6) Add the minimum uncovered element to every covered element.

22. Hungarian Algorithm(7) Subtract the minimum element from every element in the matrix.

23. Hungarian Algorithm(8) Cover the zero elements again. If the number of lines covering the zero elements is not equal to the number of rows, return to step 6.

24. Hungarian Algorithm(9) Select a matching by choosing a set of zeros so that each row or column has only one selected.

25. Hungarian Algorithm(10) Apply the matching to the original matrix, disregarding dummy rows. This shows who should do which activity, and adding the costs will give the total minimum cost.

29. Sliding window Time consuming 整段影片整段影片 時間／空間時間／空間 資料量

30. 整段影片：３ hr 切割時間／空間 (8x8) ：３ min – 時間 (8x8) – 空間 (200frame)

31. Feature

32. To do Human detection Build ground truth Post processing

33. demo

34. The end Thank you!