OpenCV 4 with Python Blueprints - Second Edition

Assembling a training dataset

Before we can train an MLP, we need to assemble a suitable training set. This is done because chances are that your face is not yet part of any dataset out there (the National Security Agency's (NSA's) private collection doesn't count), thus we will have to assemble our own. This is done most easily by returning to our GUI application from the previous chapters that can access a webcam, and operate on each frame of a video stream.

We are going to subclass the wx_gui.BaseLayout and tweak the user interface (UI) to our liking. We will have two classes for the two different modes.

The GUI will present the user with the option of recording one of the following six emotional expressions—namely, neutral, happy, sad, ...