Dlib binary CIFAR-10 dataset reading, deep learning

Question

I want to classify images in DLIB using example MNIST code: http://dlib.net/dnn_introduction_ex.cpp.html

but my dataset will be CIFAR 10, binary one: http://www.cs.toronto.edu/~kriz/cifar.html

I don't know how to make it readable and trainable for dlib.

I have found a binary reader that converts it to a struct, but still I can't make it trainable :/ https://github.com/wichtounet/cifar-10

My code:

#include "stdafx.h"
#include <cifar\cifar10_reader.hpp>
#include <dlib/dnn.h>
#include <iostream>
#include <dlib/data_io.h>


using namespace std;

using namespace dlib;

int main(int argc, char** argv) try
{

auto dataset = cifar::read_dataset<std::vector, std::vector, uint8_t, uint8_t>();

using net_type = loss_multiclass_log<
    fc<10,
    relu<fc<84,
    relu<fc<120,
    max_pool<2, 2, 2, 2, relu<con<16, 5, 5, 1, 1,
    max_pool<2, 2, 2, 2, relu<con<6, 5, 5, 1, 1,
    input<matrix<unsigned char>>
    >>>>>>>>>>>>;

net_type net;

dnn_trainer<net_type> trainer(net);
trainer.set_learning_rate(0.01);
trainer.set_min_learning_rate(0.00001);
trainer.set_mini_batch_size(128);
trainer.be_verbose();

trainer.set_synchronization_file("mnist_sync", std::chrono::seconds(20));

trainer.train(dataset.training_images, dataset.training_labels);

net.clean();
serialize("cifar.dat") << net;

std::vector<unsigned long> predicted_labels = net(dataset.training_images);
int num_right = 0;
int num_wrong = 0;

for (size_t i = 0; i < dataset.training_images.size(); ++i)
{
    if (predicted_labels[i] == dataset.training_labels[i])
        ++num_right;
    else
        ++num_wrong;

}
cout << "training num_right: " << num_right << endl;
cout << "training num_wrong: " << num_wrong << endl;
cout << "training accuracy:  " << num_right / (double)(num_right + num_wrong) << endl;


predicted_labels = net(dataset.test_images);
num_right = 0;
num_wrong = 0;
for (size_t i = 0; i < dataset.test_images.size(); ++i)
{
    if (predicted_labels[i] == dataset.test_labels[i])
        ++num_right;
    else
        ++num_wrong;

}
cout << "testing num_right: " << num_right << endl;
cout << "testing num_wrong: " << num_wrong << endl;
cout << "testing accuracy:  " << num_right / (double)(num_right + num_wrong) << endl;



net_to_xml(net, "cif.xml");
}
catch (std::exception& e)
 {
     cout << e.what() << endl;
 } 

I only replaced the original vectors as training_images to dataset.training_images. But this program doesn't compile, because of an overload in:

trainer.train(dataset.training_images, dataset.training_labels);

I'm not sure what type is dataset.training_images etc. and how to use it.