Mastering Computer Vision with TensorFlow 2.x

Selecting Haar-like features

The Haar cascade classifier algorithm is based on the idea that the image of a human face has distinctive features of intensity in different regions of the face—for example, the eye region of a face is darker than the bottom of the eyelid and the nose region is brighter than the two facial regions next to it. A Haar-like feature is represented by black and white adjacent rectangles, as shown in the following image. In this image, there are several potential Haar-like features (two-rectangle, three-rectangle, and four-rectangle features):

Note that the rectangle section is placed over the characteristic feature ...