SmartReco is a CNN (Convolutionnal Neural Network) classificator library written in C++ lang. Its goal is to identify element in pictures.
#include <SR/Image>
#include <SR/Matrix>
#include <SR/Layer>
#include <SR/Operations>
#include <SR/Data>
int main() {
SR::Network::Network net(7); // 7 Layers
SR::Data::DataSet<SQL> dataset; // Entry dataset, type : sql
SR::Network::Layer<Input> in(); // Input layer
SR::Network::Layer<Conv> conv(Relu); // Convolutional layer with ReLU correction
SR::Network::Layer<Pooling> pool(); // Pooling layer
SR::Network::Layer<FullyConnected> fc(SoftMax); // Fully Connected layer with SoftMax activation
SR::Network::Layer<Output> out(3); // Output layer
out.addType("Plane"); // First possibility of output
out.addType("Bird"); // Second possibility of output
out.addType("Other"); // Third possibility of output
SR::Data::ProbablityArray probability; // Output probability array
dataset.importFrom<Url>("https://cifar.fr/cifar.sql"); // Input origin
net.addLayers(in, conv, pool, conv, pool, fc, out); // Network construction
net.train(dataset.train); // Network training from 'dataset train set'
net.compute(dataset.test); // Network computing from 'dataset test set'
net.printResult(probability); // Network outputing to 'probability array'
return 0;
}