290 likes | 315 Views
Pattern Recognition - Face detection using Adaboosting. Presented by Master Course, 2 nd Semester Kim Sung Yong. Overview. Viola Jones technique overview Features Integral Images Feature Extraction Adaboost algorithms Cascade of boosted classifiers Training and Example Results.
E N D
Pattern Recognition - Face detection using Adaboosting Presented by Master Course, 2nd Semester Kim Sung Yong
Overview • Viola Jones technique overview • Features • Integral Images • Feature Extraction • Adaboost algorithms • Cascade of boosted classifiers • Training and Example Results
Viola Jones technique overview • Three major contributions/phases of the algorithm : • Feature extraction • Classification using boosting • Multi-scale detection algorithm • Feature extraction and feature evaluation. • Rectangular features are used, with a new image representation their calculation is very fast. • Classifier training and feature selection using a slight variation of a method called AdaBoost. • A combination of simple classifiers is very effective
Overview • Viola Jones technique overview • Features • Integral Images • Feature Extraction • Adaboost algorithms • Cascade of boosted classifiers • Training and Example Results
Features Four basic types. They are easy to calculate. The white areas are subtracted from the black ones. A special representation of the sample called the integral image makes feature extraction faster.
Overview • Viola Jones technique overview • Features • Integral Images • Feature Extraction • Adaboost algorithms • Cascade of boosted classifiers • Training and Example Results
Integral images Summed area tables A representation that means any rectangle’s values can be calculated in four accesses of the integral image.
Overview • Viola Jones technique overview • Features • Integral Images • Feature Extraction • Adaboost algorithms • Cascade of boosted classifiers • Training and Example Results
Feature Extraction Features are extracted from sub windows of a sample image. The base size for a sub window is 24 by 24 pixels. Each of the four feature types are scaled and shifted across all possible combinations In a 24 pixel by 24 pixel sub window there are ~160,000 possible features to be calculated.
Overview • Viola Jones technique overview • Features • Integral Images • Feature Extraction • Adaboost algorithms • Cascade of boosted classifiers • Training and Example Results
Recall: Perceptron Operation • Equations of “thresholded” operation: • = 1 (if w1x1 +… wd xd + wd+1 > 0) • o(x1, x2,…, xd-1, xd) • = -1 (otherwise)
Perceptron with just a Single Feature • Equations of “thresholded” operation: • Equivalent to x1 > - wd+1 / w1 i.e., equivalent to comparing the feature to a threshold Learning = finding the best threshold for a single feature Can be trained by gradient descent (or direct search) • = 1 (if w1x1 + wd+1 > 0) • o(x1) • = -1 (otherwise)
Overview • Viola Jones technique overview • Features • Integral Images • Feature Extraction • Adaboost algorithms • Cascade of boosted classifiers • Training and Example Results
Cascade of Boosted Classifiers • 단계 검사 방법 • 단계 검사 방법의 개요 • 검출 시 연산 속도를 향상하기 위해 AdaBoost 분류를 단계 구조로 만들어 적용한다. • 단계 구조는 여러 개의 스테이지(stage)를 나누고 스테이지 별로 약한 분류기의 수를 다르게 하여 수행하는 방법이다. • 처음 스테이지 분류에는 영상에 포함된 대부분의 비얼굴을 제거하는 기능이 있어 단일 스테이지를 적용하는 방법보다 좋은 성능을 얻게 된다. • 단계 검출 방법의 원리는 얼굴 영상(포지티브 영상)은 그대로 유지하면서 대부분의 비얼굴 영상(네거티브 영상)을 초기 단계에서 걸러냄으로써 나머지 단계에서 좀 더 복잡한 영상에 대한분류에 집중할 수 있도록 고안한 것이다. • 단계 검사(cascade method)첫 번째 단계에서 거짓 긍정률(false positive rate)을 어느 정도 감수해야 한다. • 높은 거짓 긍정률 자체는 분류기로 사용하기에 적합하지 않지만 이후의 단계를 거침으로써 더욱 더 판별력이 좋아지게 된다.
Overview • Viola Jones technique overview • Features • Integral Images • Feature Extraction • Adaboost algorithms • Cascade of boosted classifiers • Training and Example Results
Training In paper, 24x24 images of faces and non faces (positive and negative examples).
Sample results using the Viola-Jones Detector Notice detection at multiple scales