Machine Learning for OpenCV 4 - Second Edition

Preprocessing the data

However, since we are working with OpenCV, this time we want to make sure the input matrix is made up of 32-bit floating point numbers, otherwise the code will break:

In [2]: import numpy as np... X = X_raw.astype(np.float32)

Furthermore, we need to think back to Chapter 4, Representing Data and Engineering Features, and remember how to represent categorical variables. We need to find a way to represent target labels, not as integers but with a one-hot encoding. The easiest way to achieve this is by using scikit-learn's preprocessing module:

In [3]: from sklearn.preprocessing import OneHotEncoder...     enc = OneHotEncoder(sparse=False, dtype=np.float32)...     y = enc.fit_transform(y_raw.reshape(-1, 1))