Professional Documents
Culture Documents
Presented by
Page 1
Outline
Challenges Classification model Equipment Used Fusion-based Multiclass AdaBoost Experiment and Results Conclusion Demo
Page 2
Page 3
Guess
[1]
Page 4
Ooops.. !!
Page 5
Classification model
Feature Extraction Weak learner Boosted Classifier
Page 6
Classification model
Feature Extraction
Using useful data from input image that is unique Input image is represented in features
Page 7
Classification model
Weak learner
Gives predictions little better than random guessing Classifies using specific feature
Page 8
Classification model
Boosted Classifier
Combines the prediction of weak learner Outputs a strong hypothesis
Page 9
Classification model
Boosted Classifier Toy Example[2]
Page 10
Classification model
Boosted Classifier Toy Example Round 1
Page 11
Classification model
Boosted Classifier Toy Example Round 2
Page 12
Classification model
Boosted Classifier Toy Example Round 3
Page 13
Classification model
Boosted Classifier Toy Example Final
Page 14
Equipment Used
Thermal Camera
Captures Visual image and Thermal image
Created Chalmers Visual-Thermal database Used this database of training and testing
Page 15
Equipment Used
Thermal Camera
Captures Visual image and Thermal image
Created Chalmers Visual-Thermal database Used this database of training and testing
Masters Thesis Presentation Sep 29-2011 Page 16
Equipment Used
Thermal Camera
Captures Visual image and Thermal image
Created Chalmers Visual-Thermal database Used this database of training and testing
Masters Thesis Presentation Sep 29-2011 Page 17
Equipment Used
Thermal Camera
Captures Visual image and Thermal image
Created Chalmers Visual-Thermal database Used this database of training and testing
Page 18
Equipment Used
Thermal Camera
Captures Visual image and Thermal image
Created Chalmers Visual-Thermal database Used this database of training and testing
Page 19
Equipment Used
Chalmers Visual-Thermal database
Pose Right Front 506 Left 500 Up 456 Down 460 No. of 500 samples Cropped Visible image Cropped Thermal image
Page 20
Equipment Used
Thermal Camera
Captures Visual image and Thermal image
Created Chalmers Visual-Thermal database Used this database for training and testing
Page 21
Equipment Used
Visual and Near-Infrared Webcam
Modification of webcam to capture NIR images
Page 22
Equipment Used
Visual and Near-Infrared Webcam
Experimental Setup
Page 23
Equipment Used
Visual and Near-Infrared Webcam
Sample Images captured in DAY
Visual image
Masters Thesis Presentation Sep 29-2011
NIR image
Page 24
Equipment Used
Visual and Near-Infrared Webcam
Sample Images captured in NIGHT
Visual image
Masters Thesis Presentation Sep 29-2011
NIR image
Page 25
Equipment Used
Microsoft Kinect
Captures Visual, NIR and depth image
Page 26
Equipment Used
Microsoft Kinect
Captures Visual, NIR and depth image
Page 27
Page 28
Page 29
Page 30
Page 31
Page 32
Page 33
Page 34
Frequency
40 30 20 10 0 -15
-10
-5
10
15
Page 37
Update weights
For the above samples
Weight the weak hypothesis,
Masters Thesis Presentation Sep 29-2011 Page 38
End Loop
End Loop
Masters Thesis Presentation Sep 29-2011 Page 39
=
=1 =1
= ,
= 1, . . ,
E = max
Masters Thesis Presentation Sep 29-2011 Page 40
Page 41
Page 42
Page 43
Output:
= max
Page 44
10
15 20 Boosting Iterations
25
30
Page 45
10
15 20 Boosting Iterations
25
30
Page 46
10
15 20 Boosting Iterations
25
30
Page 47
Page 48
0.08
0.06
0.04
0.02
Page 49
Page 50
Conclusion
Fused-based Multiclass AdaBoost was proposed Fusion improves the performance of classification
Classification rate was improved from 95% to 99%
Future work
Images to Video Classification to tracking
Masters Thesis Presentation Sep 29-2011 Page 51
Demo
Any Volunteer ..?
Page 52
Questions?
Page 53
Reference
[1] Antonio Torralba, Understanding Visual Scenes,NIPS09 Computer Science and Artificial Intelligence Laboratory (CSAIL) Department of Electrical Engineering and Computer Science [2] Rob Schapire, Theory and Applications of Boosting, NIPS09
Page 54