Implementing an MLP in OpenCV uses the same syntax that we have seen at least a dozen times before. In order to see how an MLP compares to a single perceptron, we will operate on the same toy data as before:

In [1]: from sklearn.datasets.samples_generator import make_blobs
...     X_raw, y_raw = make_blobs(n_samples=100, centers=2,
...                               cluster_std=5.2, random_state=42)