NOTE: Make sure the DLL library is in the same folder as the .NET executable
program is. This code was tested using .NET 8 but it may be backwards
compatible. The code requires the NuGet package SixLabors.ImageSharp v3.1.4.
Notice that the code requires some command-line arguments, so use them.

Some important notes, the images it will read into the CNN are categorized
a.k.a. classified by the subfolder they are in. So images should be placed
into subfolder with the name of their label. For example, cats should go into
a folder named cat and dogs should go into a folder named dog, and the parent
folder of these subfolders should be the directory you pass to the command
line of the console application.

Depending on the type of images, the CNN may need to be adjusted to use
different filters a.k.a. kernels.

The C++ code is only about 15% faster than the C# .NET 8 code! But, it is
slightly more efficient with memory.

Do not forget the ConfusionMatrix.cs, CCE-NeuralNetwork-Exports.cs,
ConvolutionalNeuralNetwork.cs and CceNeuralNetwork.cs files.

FYI, SoftMax has a capital M.