Pedestrian Detection and Localization
description
Transcript of Pedestrian Detection and Localization
![Page 1: Pedestrian Detection and Localization](https://reader035.fdocuments.net/reader035/viewer/2022062310/568160c6550346895dcff437/html5/thumbnails/1.jpg)
LOGOPedestrian Detection and Localization
Members:Đặng Trương Khánh Linh 0612743
Bùi Huỳnh Lam Bửu 0612733
Advisor: A.Professor Lê Hoài Bắc
UNIVERSITY OF SCIENCE ADVANCED PROGRAM IN COMPUTER SCIENCE
Year 20111
![Page 2: Pedestrian Detection and Localization](https://reader035.fdocuments.net/reader035/viewer/2022062310/568160c6550346895dcff437/html5/thumbnails/2.jpg)
LOGOProblem statement
Build up a system which automatically detects and localizes pedestrians in static image.
2
![Page 3: Pedestrian Detection and Localization](https://reader035.fdocuments.net/reader035/viewer/2022062310/568160c6550346895dcff437/html5/thumbnails/3.jpg)
LOGOApplications
Using in smart car system, or smart camera in general.
Build a software to categorize personal album images to proper catalogue.
Video tracking smart surveillanceAction recg
3
![Page 4: Pedestrian Detection and Localization](https://reader035.fdocuments.net/reader035/viewer/2022062310/568160c6550346895dcff437/html5/thumbnails/4.jpg)
LOGOChallenges
Huge variation in intra-class.Variable appearance and clothing.Complex background.Non-constraints illumination.Occlusions, different scales.
4
![Page 5: Pedestrian Detection and Localization](https://reader035.fdocuments.net/reader035/viewer/2022062310/568160c6550346895dcff437/html5/thumbnails/5.jpg)
LOGOOutlineExisting approaches.Motivation.Overview of methodology.
Learning phase Detection
Some contributions: Spatial selective approach Multi-level based approach
Non-maxima SuppressionConclusionsFuture workReference
5
![Page 6: Pedestrian Detection and Localization](https://reader035.fdocuments.net/reader035/viewer/2022062310/568160c6550346895dcff437/html5/thumbnails/6.jpg)
LOGOExisting approaches
Haar wavelets + SVM: Papageorgiou & Poggio, 2000; Mohan et al 2000
Rectangular differential features + adaBoost: Viola & Jones, 2001
Model based methods: Felzenszwalb & Huttenlocher, 2000; Loffe & Forsyth, 1999
Lowe, 1999 (SIFT).LBP, HOG, …
6
![Page 7: Pedestrian Detection and Localization](https://reader035.fdocuments.net/reader035/viewer/2022062310/568160c6550346895dcff437/html5/thumbnails/7.jpg)
LOGOHOG
7
![Page 8: Pedestrian Detection and Localization](https://reader035.fdocuments.net/reader035/viewer/2022062310/568160c6550346895dcff437/html5/thumbnails/8.jpg)
LOGOMotivation of choosing HOG
The blob structure based methods are false to object detection problem.
Use the advantage of rigid shape of object.Low complexity and fast running time.Has a good performance.
8
![Page 9: Pedestrian Detection and Localization](https://reader035.fdocuments.net/reader035/viewer/2022062310/568160c6550346895dcff437/html5/thumbnails/9.jpg)
LOGOContributions
Re-implement HOG-based pedestrian detector.Spatial Selective Method.Multi-level Method.
9
![Page 10: Pedestrian Detection and Localization](https://reader035.fdocuments.net/reader035/viewer/2022062310/568160c6550346895dcff437/html5/thumbnails/10.jpg)
LOGOOverview of methodology
10
![Page 11: Pedestrian Detection and Localization](https://reader035.fdocuments.net/reader035/viewer/2022062310/568160c6550346895dcff437/html5/thumbnails/11.jpg)
LOGOLearning Phase
11
![Page 12: Pedestrian Detection and Localization](https://reader035.fdocuments.net/reader035/viewer/2022062310/568160c6550346895dcff437/html5/thumbnails/12.jpg)
LOGODetection Phase
12
![Page 13: Pedestrian Detection and Localization](https://reader035.fdocuments.net/reader035/viewer/2022062310/568160c6550346895dcff437/html5/thumbnails/13.jpg)
LOGOScan image at all positions and scales
Object/Non-object
classifier
13
![Page 14: Pedestrian Detection and Localization](https://reader035.fdocuments.net/reader035/viewer/2022062310/568160c6550346895dcff437/html5/thumbnails/14.jpg)
LOGOContributions
IMPROVEMENTS
PERFORMANCE(Spatial Selective Approach)
ACCURACY(Multi-Level Approach)
14
![Page 15: Pedestrian Detection and Localization](https://reader035.fdocuments.net/reader035/viewer/2022062310/568160c6550346895dcff437/html5/thumbnails/15.jpg)
LOGOSpatial Selective Approach
Less informative region
[A1,..,Z1] [A2,..,Z2] [A3,..,Z3] [A4,..,Z4]
[A1,..,Z1, A2,..,Z2, A3,..,Z3, A4,..,Z4]
Descriptor
15
![Page 16: Pedestrian Detection and Localization](https://reader035.fdocuments.net/reader035/viewer/2022062310/568160c6550346895dcff437/html5/thumbnails/16.jpg)
LOGOSpatial Selective Approach
Examples:
16
![Page 17: Pedestrian Detection and Localization](https://reader035.fdocuments.net/reader035/viewer/2022062310/568160c6550346895dcff437/html5/thumbnails/17.jpg)
LOGOMulti-level Approach
Purpose: enhance the performance by getting more information about shape and contour of object.
Diff distance
17
![Page 18: Pedestrian Detection and Localization](https://reader035.fdocuments.net/reader035/viewer/2022062310/568160c6550346895dcff437/html5/thumbnails/18.jpg)
LOGOMulti-level Approach
[A1,..,Z1] [A2,..,Z2] [A3,..,Z3]
[A1,..,Z1, A2,..,Z2, A3,..,Z3, A4,..,Z4]
HOG
18
![Page 19: Pedestrian Detection and Localization](https://reader035.fdocuments.net/reader035/viewer/2022062310/568160c6550346895dcff437/html5/thumbnails/19.jpg)
LOGOMean Shift as Non-maxima Suppression
19
![Page 20: Pedestrian Detection and Localization](https://reader035.fdocuments.net/reader035/viewer/2022062310/568160c6550346895dcff437/html5/thumbnails/20.jpg)
LOGO
Region ofinterest
Center ofmass
Mean Shiftvector
Mean shift
Slide by Y. Ukrainitz & B. Sarel20
![Page 21: Pedestrian Detection and Localization](https://reader035.fdocuments.net/reader035/viewer/2022062310/568160c6550346895dcff437/html5/thumbnails/21.jpg)
LOGO
Region ofinterest
Center ofmass
Mean Shiftvector
Mean shift
Slide by Y. Ukrainitz & B. Sarel21
![Page 22: Pedestrian Detection and Localization](https://reader035.fdocuments.net/reader035/viewer/2022062310/568160c6550346895dcff437/html5/thumbnails/22.jpg)
LOGO
Region ofinterest
Center ofmass
Mean Shiftvector
Mean shift
Slide by Y. Ukrainitz & B. Sarel22
![Page 23: Pedestrian Detection and Localization](https://reader035.fdocuments.net/reader035/viewer/2022062310/568160c6550346895dcff437/html5/thumbnails/23.jpg)
LOGO
Region ofinterest
Center ofmass
Mean Shiftvector
Mean shift
Slide by Y. Ukrainitz & B. Sarel23
![Page 24: Pedestrian Detection and Localization](https://reader035.fdocuments.net/reader035/viewer/2022062310/568160c6550346895dcff437/html5/thumbnails/24.jpg)
LOGO
Region ofinterest
Center ofmass
Mean Shiftvector
Mean shift
Slide by Y. Ukrainitz & B. Sarel24
![Page 25: Pedestrian Detection and Localization](https://reader035.fdocuments.net/reader035/viewer/2022062310/568160c6550346895dcff437/html5/thumbnails/25.jpg)
LOGO
Region ofinterest
Center ofmass
Mean Shiftvector
Mean shift
Slide by Y. Ukrainitz & B. Sarel25
![Page 26: Pedestrian Detection and Localization](https://reader035.fdocuments.net/reader035/viewer/2022062310/568160c6550346895dcff437/html5/thumbnails/26.jpg)
LOGO
Region ofinterest
Center ofmass
Mean shift
Slide by Y. Ukrainitz & B. Sarel26
![Page 27: Pedestrian Detection and Localization](https://reader035.fdocuments.net/reader035/viewer/2022062310/568160c6550346895dcff437/html5/thumbnails/27.jpg)
LOGOMean shift clustering
• Cluster: all data points in the attraction basin of a mode
• Attraction basin: the region for which all trajectories lead to the same mode
Slide by Y. Ukrainitz & B. Sarel27
![Page 28: Pedestrian Detection and Localization](https://reader035.fdocuments.net/reader035/viewer/2022062310/568160c6550346895dcff437/html5/thumbnails/28.jpg)
LOGONon-maximum suppression
Using non-maximum suppression such as mean shift to find the modes.
28
![Page 29: Pedestrian Detection and Localization](https://reader035.fdocuments.net/reader035/viewer/2022062310/568160c6550346895dcff437/html5/thumbnails/29.jpg)
LOGOExperiments
Re-implement Dalal‘s MethodSpatial Selective MethodMulti-Level Method
29
![Page 30: Pedestrian Detection and Localization](https://reader035.fdocuments.net/reader035/viewer/2022062310/568160c6550346895dcff437/html5/thumbnails/30.jpg)
LOGOResult of experiment
30
![Page 31: Pedestrian Detection and Localization](https://reader035.fdocuments.net/reader035/viewer/2022062310/568160c6550346895dcff437/html5/thumbnails/31.jpg)
LOGOResult Spatial Selective Method
31
![Page 32: Pedestrian Detection and Localization](https://reader035.fdocuments.net/reader035/viewer/2022062310/568160c6550346895dcff437/html5/thumbnails/32.jpg)
LOGOVector Length v.s Speed
Origin 1:0 1:1 1.5:0 1.5:10
500
1000
1500
2000
2500
3000
3500
4000
0
1
2
3
4
5
6
Vector lengthSpeed (in minutes)
A:B Deleted cell(s): Overlap cell(s)32
![Page 33: Pedestrian Detection and Localization](https://reader035.fdocuments.net/reader035/viewer/2022062310/568160c6550346895dcff437/html5/thumbnails/33.jpg)
LOGOResult Multi-Level
33
![Page 34: Pedestrian Detection and Localization](https://reader035.fdocuments.net/reader035/viewer/2022062310/568160c6550346895dcff437/html5/thumbnails/34.jpg)
LOGOResult(cont…)
34
![Page 35: Pedestrian Detection and Localization](https://reader035.fdocuments.net/reader035/viewer/2022062310/568160c6550346895dcff437/html5/thumbnails/35.jpg)
LOGODataset
INRIA pedestrian dataset
Train:1208 positive windows1218 negative images
Test:566 positive windows453 negative images
35
![Page 36: Pedestrian Detection and Localization](https://reader035.fdocuments.net/reader035/viewer/2022062310/568160c6550346895dcff437/html5/thumbnails/36.jpg)
LOGOConclusions
Successfully re-implement HOG descriptor.Propose the Spatial Selective Approach which take
advantages of less informative center region of image window.
Multi-level has more information about shape and contour of object.
36
![Page 37: Pedestrian Detection and Localization](https://reader035.fdocuments.net/reader035/viewer/2022062310/568160c6550346895dcff437/html5/thumbnails/37.jpg)
LOGOFuture work
Non-uniform grid of points.Combination of Spatial Selective and Multi-level
approach.
37
![Page 38: Pedestrian Detection and Localization](https://reader035.fdocuments.net/reader035/viewer/2022062310/568160c6550346895dcff437/html5/thumbnails/38.jpg)
LOGONon-uniform grid of points
38
![Page 39: Pedestrian Detection and Localization](https://reader035.fdocuments.net/reader035/viewer/2022062310/568160c6550346895dcff437/html5/thumbnails/39.jpg)
LOGOReferences N. Dalal and B. Triggs, “Histograms of oriented
gradients for human detection,” in IEEE Conference on Computer Vision and Pattern Recognition, 2005.
Subhransu Maji et al. Classification using Intersection Kernel Support Vector Machines is Efficient. IEEE Computer Vision and Pattern Recognition 2008
C. Harris and M. Stephens. A combined corner and edge detector. In Alvey Vision Conference, pages 147–151, 1988.
D. G. Lowe. Distinctive image features from scale-invariant keypoints. International Journal of Computer Vision, 60(2):91–110, 2004.
39