Learning OpenCV 4 Computer Vision with Python 3 - Third Edition

Trying to improve the ANN's training

We could apply a number of potential improvements to the problem of training our ANN. We have already mentioned some of these potential improvements, but let's review them here:

• You could experiment with the size of your training dataset, the number of hidden nodes, and the number of epochs until you find a peak level of accuracy.
• You could modify our digits_ann.create_ann function so that it supports more than one hidden layer.
• You could also try different activation functions. We have used cv2.ml.ANN_MLP_SIGMOID_SYM, but it isn't the only option; the others include cv2.ml.ANN_MLP_IDENTITY, cv2.ml.ANN_MLP_GAUSSIAN, cv2.ml.ANN_MLP_RELU, and cv2.ml.ANN_MLP_LEAKYRELU.
• Similarly, you could try different ...