1. Robust Real-Time Object Detection Paul Viola & Michael Jones

2. Introduction Frontal face detection is achieved Comparatively satisfactory detection rates Efficient decrease in false positive rate Extremely rapid operation 384*288 pixel image is processed for 15 frames/second

3. Contribution of The Paper Integral image A new image representation AdaBoost Effective classifier selection Cascade structure of complex classifiers Dramatic decrease in detection time

4. Simple Rectangle Features Why not use pixels directly ? Features encodes domain knowledge that is hard to learn by finite quantity of training data Features operates much faster than pixel based systems

5. Integral Image Double integral of original image A new representation of image for fast calculation of rectangle features

6. Integral Image Sum of pixels in rectangle D from the original image can be defined in integral image as : P(4) -P(3)-P(2)+P(1)

7. Advantages of Integral Image Pyramid image Requires a pyramid of images A fixed scaled detector works on all those images Forming the pyramid is computationally expensive Integral Image A single feature can be evaluated at any scale and location in a few operations Integral image is computed in one pass over the original image

8. Learning Classification Functions 45,394 features associated with each sub- window A very small number of these features can be combined to form an effective classifier A variant of AdaBoost is used to Select features Train the classifier

9. How does AdaBoost work? Combines a mixture of weak classifiers to form a strong one Percepton algorithm returns the one having the minimum classification error The examples are re-weighted in according to the accuracy of the first classifier The final strong classifier is a weighted combination of weak classifiers

10. How does AdaBoost work? First and second features selected by AdaBoost

11. Attentional Cascade Increase detection performance & reduce computation time Calling simpler classifiers before complex ones A simple classifier example (two-feature): 100% detection rate 40% false positives 60 microprocessor instructions (very efficient)

12. Attentional Cascade

13. Training of Cascade of Classifiers The deeper classifiers are trained with harder examples Simple classifiers in the first stages, complex ones in the deeper parts of the cascade Complex classifiers takes more time to compute

14. A general detection algorithm works like 85-95% detection rate & 10 -5 - 10 -6 % false positive rate The cascade system works like With a 10 stage classifier For each cascade a detection rate 99% and false positive rate 30 % Overall system runs at (.99 10 ~) 90% detection rate (0.30 10 ~)6 * 10 -6 % false positive rate Training of Cascade of Classifiers

15. Requirements Needs to be determined : Number of stages Number of features for each stage Thresholds for each stage

16. Practical Implementation User selects acceptable f i and d i for each layer Each layer is trained for Adaboost Number of features are increased until target f i and d i are met for this level If overall target F and D is not met for the system we add a new level to the cascade

18. Results – Structure of Cascade 32 layers – 4297 features Weeks spent to train the cascade Layer #1234567… Features2520 50 … False Positives40%20% Detection Rate100%

19. Results – Algorithm Details All sub-windows (training – testing) are variance normalized for lighting conditions Scaling is achieved by just scaling the detectors rather than the image Step size of one pixel is used

20. Results – Algorithm details

21. Results Most of the windows are rejected in the first & second cascade Face detection on a 384x288 image runs in about 0.067 seconds 15 times faster than Rowley-Baluja-Kanade 600 times faster than Schneiderman- Kanade

22. Results 150 images and 507 faces