テレビ映像を指向したJoint Cascade 顔検出の学習と評価

抄録

Face detection is one of the most important techniques for video content analysis. The joint cascade method[2] enables fast and accurate face detection by combining face alignment with detection. We train and evaluate a classifier of the joint cascade to investigate an ability for TV video analysis. In experiments, we used 20000 face images to train the classifier and evaluated its effectiveness.