Presentation is loading. Please wait.

Presentation is loading. Please wait.

Tracking Learning Detection

Similar presentations


Presentation on theme: "Tracking Learning Detection"— Presentation transcript:

1 Tracking Learning Detection
A Camera that Learns.

2 Why is object tracking so important ?
vehicle detection motion capture robot vision Why is object tracking so important ? scene analysis hand recognition surveillance

3

4 What is TLD intuitively?
Outline Why TLD? - Traditional methods - Challenges to old ways What is TLD intuitively? - Framework - Key ideas How does TLD work exactly? - Tracking - Learning - Detection - TLD Experiment Results

5 Why TLD ?

6 Why TLD ? 你會怎麼做?

7 Why TLD ? - Traditional method 1 :
Tracking Estimate motion over frames Only need initialization. Appearance changing & Partial occlusions. (?) Accumulate error. Fail if the object disappears.

8 Why TLD ? - Traditional method 2 :
Detection Off-line Learning Scan every frame independently Do not fail if the object disappears. Need off-line training. Appearance changing & Partial occlusions.

9 Why TLD ? - Challenges to old ways
Not fail if the object disappears. Appearance changing. Partial occlusions.

10 Why TLD ? - Challenges to old ways
Not fail if the object disappears. TLD is an algorithm that can solve all these tasks ! Appearance changing. Partial occlusions.

11 What is TLD intuitively?

12 What is TLD intuitively? - Framework
Tracking Off-line Learning Detection

13 What is TLD intuitively? - Framework
On-line Learning Trajectory fragment Location Update Tracking Detection Update

14 What is TLD intuitively? - Key ideas 1 : Learning
Updates training examples (add/delete) Detect Track & detect Update detector Appearance changing. Track

15 What is TLD intuitively? - Key ideas 2 : Updating
True trajectory Tracking Detection TLD Not fail if no object. Partial occlusions.

16 How does TLD work exactly?

17 How does TLD work exactly? - Framework

18 How does TLD work exactly? - Tracking
1.Reliable point selection: - Forward-backward selection Choose point 1 Discard point 2 - Use forward-backward selection to select 50% of the most reliable points.

19 How does TLD work exactly? - Tracking
2.Median-Flow tracker: and estimates displacement using median with confidence proportional to (1/scale of the median). Tracks the reliable points between consecutive frames ,

20 How does TLD work exactly? - Framework

21 How does TLD work exactly? - Detection
2bitBP features Scanning window 2BitBP : invariance to illumination RF : fast Illumination invariance 2bitBP features

22 How does TLD work exactly? - Detection
Random Forest Random Forest Posteriors Scanning window 2BitBP : invariance to illumination RF : fast If the average posteriors is larger than 50%, consider it an object 2bitBP features

23 How does TLD work exactly? - Detection
1-NN classifier Confidence =d-/(d-+d+) Feature space Random Forest Posteriors Scanning window 2bitBP features

24 How does TLD work exactly? - Framework
Pick the one with higher confidence score

25 How does TLD work exactly? - Learning
Feature space

26 How does TLD work exactly? - Learning
Updates training examples (add/delete) Detect Track & detect Update detector Track

27 How does TLD work exactly? - Learning :
Appearance changing problem solved! 1 6 10 9

28 How does TLD work exactly? - Recap : t=0

29 How does TLD work exactly? - Recap : t=1
Decide by confidence score

30 How does TLD work exactly? - Recap : t=1

31 How does TLD work exactly? - Recap : t=2
Remove if false detect

32 Experiment Result

33 Experiment Result More vid

34 Experiment Result

35 Experiment Result

36 Thank you


Download ppt "Tracking Learning Detection"

Similar presentations


Ads by Google