推薦論文:http://cs229.stanford.edu/proj2008/Jo-FaceDetectionUsingLBPfeatures.pdf
注:實現的是人臉識別(face detection), 而不是人臉檢測(face recognization), 意思是不一樣的!不一樣的!不一樣的!
前一篇文章 我們已經利用LBP得到特徵,那麼就要對這些特徵進行“手術”了!
完整的三步驟:
1. extract features from the images using LBP;
2. AdaBoost(recommend Gentle AdaBoost) is used to select features and to train the classifier.
3. To compose a strong classifier, a cascade of classifiers is used!
yes, the classifier is what we want. Getting fevered now! why? because we are familiar with classifier!
by the way, a liitle supplement, in procedure 2, the training set includes/adds rotated faces and in low light conditions' faces.
remember this phrase, LBP-based classifier.
容我靜靜地裝逼,這東西英文寫的準確,捂臉。
用到的知識很簡單, 就是Adaboost算法和級聯分類器的知識!easy!
稍後代碼放上來。
https://github.com/GitHubGS/Face-Recognition-and-Detection-Analysis