Human-Computer Interaction Sergey Kosov
Lecture 3: Face Detection
Contents
1. Introduction to Face Detection
2. Image features
3. Boosting
4. Attentional cascade
5. Implementation and training
© 2021 Sergey Kosov
1
Face detection
2
Face detection
Basic idea:
slide a window across image and evaluate a face model at every location
3
Challenges of face detection
Sliding window detector must evaluate tens of thousands of location / scale combinations
Faces are rare: 0 – 10 per image
• For computational efficiency, we should try to spend as little time as possible on the non-face windows
• A megapixel image has ~106 pixels and a comparable number of candidate face locations
• To avoid having a false positive in every image, our false positive rate has to be less than 10-6
4
The Viola / Jones Face Detector
A seminal approach to real-time object detection
Training is slow, but detection is very fast
Key ideas
• Integral images for fast feature evaluation
• Boosting for feature selection
• Attentional cascade for fast rejection of non-face windows
P. Viola and M. Jones: Rapid object detection using a boosted cascade of simple features. CVPR 2001. P. Viola and M. Jones: Robust real-time face detection. IJCV 57(2), 2004.
5
Image Features
“Rectangle filters”
Value = ∑ (pixels in white area) – ∑ (pixels in black area)
6
Example
Source
Result
7
Fast computation with integral images
The integral image
computes a value at each pixel (𝑥, 𝑦) that is the sum of the pixel values above and to the left of (𝑥, 𝑦), inclusive
This can quickly be computed in one pass through the image
(𝑥, 𝑦)
8
Computing the integral image
9
Computing the integral image
𝑖𝑖(𝑥, 𝑦 − 1)
𝑠(𝑥 − 1, 𝑦)
𝑖(𝑥, 𝑦)
Cumulativerowsum:𝑠(𝑥,𝑦) = 𝑠(𝑥–1,𝑦) + 𝑖(𝑥,𝑦) Integralimage:𝑖𝑖(𝑥,𝑦) = 𝑖𝑖(𝑥,𝑦−1) + 𝑠(𝑥,𝑦)
10
Computing sum within a rectangle
Let A,B,C,D be the values of the integral image at the corners of a rectangle
Then the sum of original image values within the rectangle can be computed as:
• sum=A–B–C+D
Only 3 additions are required for any size of rectangle!
DB
CA
11
Example
Integral Image
+1 +2 -2
-1
-1 +1
12
Feature selection
For a 24×24 detection region, the number of possible rectangle features is ~160,000!
13
Feature selection
For a 24×24 detection region, the number of possible rectangle features is ~160,000!
At test time, it is impractical to evaluate the entire feature set
Can we create a good classifier using just a small subset of all possible features?
How to select such a subset?
14
Boosting
Boosting is a classification scheme that works by combining weak learners into a more accurate ensemble classifier
• A weak learner need only do better than chance Training consists of multiple boosting rounds
• During each boosting round, we select a weak learner that does well on examples that were hard for the previous weak learners
• “Hardness” is captured by weights attached to training examples
Y. Freund and R. Schapire, A short introduction to boosting, Journal of Japanese Society for Artificial Intelligence, 14(5):771-780, September, 1999.
15
Training procedure
Initially, weight each training example equally In each boosting round:
• Find the weak learner that achieves the lowest weighted training error
• Raise the weights of training examples misclassified by current weak learner
Compute final classifier as linear combination of all weak learners (weight of each learner is directly proportional to its accuracy)
Exact formulas for re-weighting and combining weak learners depend on the particular boosting scheme (e.g., AdaBoost)
Y. Freund and R. Schapire, A short introduction to boosting, Journal of Japanese Society for Artificial Intelligence, 14(5):771-780, September, 1999.
16
Boosting illustration
Weak Classifier 1
17
Boosting illustration
Weights Increased
18
Boosting illustration
Weak Classifier 2
19
Boosting illustration
Weights Increased
20
Boosting illustration
Weak Classifier 3
21
Boosting illustration
Final classifier is
a combination of weak classifiers
22
Boosting vs. SVM
Advantages of boosting
• Integrates classification with feature selection
• Complexity of training is linear instead of quadratic in the number of training examples
• Flexibility in the choice of weak learners, boosting scheme
• Testing is fast
• Easy to implement
Disadvantages
• Needs many training examples
• Often doesn’t work as well as SVM (especially for many-class problems)
23
Boosting for face detection
Define weak learners based on rectangle features
value of rectangle feature
1 ifpf(x)>pq h(x)=ìítt tt
t î0 otherwise parity threshold window
24
Boosting for face detection
Define weak learners based on rectangle features For each round of boosting:
• Evaluate each rectangle filter on each example
• Select best threshold for each filter
• Select best filter / threshold combination
• Reweight examples
Computational complexity of learning: O(MNK)
• M rounds, N examples, K features
25
Boosting for face detection
First two features selected by boosting:
This feature combination can yield 100% detection rate and 50% false positive rate
26
Boosting for face detection
A 200-feature classifier can yield 95% detection rate and a false positive rate of 1 in 14084
Not good enough!
Receiver operating characteristic (ROC) curve
27
Attentional cascade
We start with simple classifiers which reject many of the negative sub- windows while detecting almost all positive sub-windows
Positive response from the first classifier triggers the evaluation of a second (more complex) classifier, and so on
A negative outcome at any point leads to the immediate rejection of the sub-window
IMAGE SUB-WINDOW
Classifier 1
T T T Classifier 2 Classifier 3
FACE
FFF NON-FACE NON-FACE NON-FACE
28
Attentional cascade
Chain classifiers that are progressively more complex and have lower false
positive rates:
Receiver operating characteristic
% False Pos
0 50
vs false negdetermined by
IMAGE SUB-WINDOW
Classifier 1
T T T Classifier 2 Classifier 3
FACE
FFF NON-FACE NON-FACE NON-FACE
29
0
100
% Detection
Attentional cascade
The detection rate and the false positive rate of the cascade are found by multiplying the respective rates of the individual stages
A detection rate of 0.9 and a false positive rate on the order of 10-6 can be achieved by a 10-stage cascade if each stage has a detection rate of 0.99 (0.9910 ≈ 0.9) and a false positive rate of about 0.30 (0.310 ≈ 6×10-6)
IMAGE SUB-WINDOW
Classifier 1
T T T Classifier 2 Classifier 3
FACE
FFF NON-FACE NON-FACE NON-FACE
30
Training the cascade
Set target detection and false positive rates for each stage
Keep adding features to the current stage until its target rates have been met
• Need to lower AdaBoost threshold to maximize detection (as opposed to minimizing total classification error)
• Test on a validation set
If the overall false positive rate is not low enough, then add another stage
Use false positives from current stage as the negative training examples for the next stage
31
Training Data
• 5000 faces
• All frontal, rescaled to
24×24 pixels
• 300 million non-faces
• 9500 non-face images • Faces are normalized
• Scale, translation Many variations
• Across individuals
• Illumination
• Pose
The implemented system
32
System performance
Training time: “weeks” on 466 MHz Sun workstation
38 layers, total of 6061 features
Average of 10 features evaluated per window on test set
“On a 700 Mhz Pentium III processor, the face detector can process a 384 by 288 pixel image in about .067 seconds”
• 15Hz
• 15 times faster than previous detector of comparable accuracy (Rowley et al., 1998)
33
Output of Face Detector on Test Images
34
Other detection tasks
Facial Feature Localization Profile Detection
Male vs. female
35
Profile Detection
36
Profile Features
37
Wrap – Up
Rectangle features
Integral images for fast computation
Boosting for feature selection
Attentional cascade for fast rejection of negative windows
38
Assignment 1
Submission deadline: Friday, 26. February 2020 23:55
Programming Assignments should be submitted via pull request on GitHub. Every assignment sheets counts 100 points.
1.1 Implementation of a Minimal Face Recognition System (30 + 70 Points)
• Goto: https://github.com/Jacobs-University/visir-traker-01 (for the task details)
Need Help? Ask TAs! and / or join HCI Slack workspace!
• To join the HCI Slack workspace, please write me at s.kosov@jacobs-university.de
a short (or empty) e-mail with subject: “Join CG SLACK”. I will reply with email invitation.
Assistants
• Anish Ghosh
• Email: ani.ghosh@jacobs-university.de
39