Machine Learning for OpenCV 4 - Second Edition

Knowing our kernels

OpenCV provides a whole range of SVM kernels to experiment with. Some of the most commonly used ones include the following:

• cv2.ml.SVM_LINEAR: This is the kernel we used previously. It provides a linear decision boundary in the original feature space (the x and y values).
• cv2.ml.SVM_POLY: This kernel provides a decision boundary that is a polynomial function in the original feature space. In order to use this kernel, we also have to specify a coefficient via svm.setCoef0 (usually set to 0) and the degree of the polynomial via svm.setDegree.
• cv2.ml.SVM_RBF: This kernel implements the kind of Gaussian function we discussed earlier.
• cv2.ml.SVM_SIGMOID: This kernel implements a sigmoid function, similar to the one we encountered ...