2016-06-23 10:26:48 +08:00
|
|
|
// The contents of this file are in the public domain. See LICENSE_FOR_EXAMPLE_PROGRAMS.txt
|
|
|
|
/*
|
2016-06-25 21:26:51 +08:00
|
|
|
This example shows how to classify an image into one of the 1000 imagenet
|
|
|
|
categories using the deep learning tools from the dlib C++ Library. We will
|
|
|
|
use the pretrained ResNet34 model available on the dlib website.
|
2016-06-23 10:26:48 +08:00
|
|
|
|
2016-06-25 21:26:51 +08:00
|
|
|
The ResNet34 architecture is from the paper Deep Residual Learning for Image
|
|
|
|
Recognition by He, Zhang, Ren, and Sun. The model file that comes with dlib
|
|
|
|
was trained using the dnn_imagenet_train_ex.cpp program on a Titan X for
|
|
|
|
about 2 weeks. This pretrained model has a top5 error of 7.572% on the 2012
|
|
|
|
imagenet validation dataset.
|
|
|
|
|
2016-06-25 21:40:11 +08:00
|
|
|
For an introduction to dlib's DNN module read the dnn_introduction_ex.cpp and
|
|
|
|
dnn_introduction2_ex.cpp example programs.
|
2016-06-23 10:30:45 +08:00
|
|
|
|
2016-06-23 10:26:48 +08:00
|
|
|
|
2016-06-25 21:26:51 +08:00
|
|
|
Finally, these tools will use CUDA and cuDNN to drastically accelerate
|
|
|
|
network training and testing. CMake should automatically find them if they
|
|
|
|
are installed and configure things appropriately. If not, the program will
|
2016-06-23 10:26:48 +08:00
|
|
|
still run but will be much slower to execute.
|
|
|
|
*/
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
#include <dlib/dnn.h>
|
|
|
|
#include <iostream>
|
|
|
|
#include <dlib/data_io.h>
|
|
|
|
#include <dlib/gui_widgets.h>
|
|
|
|
#include <dlib/image_transforms.h>
|
|
|
|
|
|
|
|
using namespace std;
|
|
|
|
using namespace dlib;
|
|
|
|
|
|
|
|
// ----------------------------------------------------------------------------------------
|
|
|
|
|
2016-06-25 21:26:51 +08:00
|
|
|
// This block of statements defines the resnet-34 network
|
2016-06-23 10:26:48 +08:00
|
|
|
|
|
|
|
template <template <int,template<typename>class,int,typename> class block, int N, template<typename>class BN, typename SUBNET>
|
|
|
|
using residual = add_prev1<block<N,BN,1,tag1<SUBNET>>>;
|
|
|
|
|
|
|
|
template <template <int,template<typename>class,int,typename> class block, int N, template<typename>class BN, typename SUBNET>
|
|
|
|
using residual_down = add_prev2<avg_pool<2,2,2,2,skip1<tag2<block<N,BN,2,tag1<SUBNET>>>>>>;
|
|
|
|
|
|
|
|
template <int N, template <typename> class BN, int stride, typename SUBNET>
|
|
|
|
using block = BN<con<N,3,3,1,1,relu<BN<con<N,3,3,stride,stride,SUBNET>>>>>;
|
|
|
|
|
|
|
|
template <int N, typename SUBNET> using ares = relu<residual<block,N,affine,SUBNET>>;
|
|
|
|
template <int N, typename SUBNET> using ares_down = relu<residual_down<block,N,affine,SUBNET>>;
|
|
|
|
|
2016-10-09 02:27:34 +08:00
|
|
|
template <typename SUBNET> using level1 = ares<512,ares<512,ares_down<512,SUBNET>>>;
|
|
|
|
template <typename SUBNET> using level2 = ares<256,ares<256,ares<256,ares<256,ares<256,ares_down<256,SUBNET>>>>>>;
|
|
|
|
template <typename SUBNET> using level3 = ares<128,ares<128,ares<128,ares_down<128,SUBNET>>>>;
|
|
|
|
template <typename SUBNET> using level4 = ares<64,ares<64,ares<64,SUBNET>>>;
|
2016-06-23 10:26:48 +08:00
|
|
|
|
2016-06-25 21:26:51 +08:00
|
|
|
using anet_type = loss_multiclass_log<fc<1000,avg_pool_everything<
|
2016-10-09 02:27:34 +08:00
|
|
|
level1<
|
|
|
|
level2<
|
|
|
|
level3<
|
|
|
|
level4<
|
2016-06-23 10:26:48 +08:00
|
|
|
max_pool<3,3,2,2,relu<affine<con<64,7,7,2,2,
|
|
|
|
input_rgb_image_sized<227>
|
2016-10-09 02:27:34 +08:00
|
|
|
>>>>>>>>>>>;
|
2016-06-23 10:26:48 +08:00
|
|
|
|
|
|
|
// ----------------------------------------------------------------------------------------
|
|
|
|
|
|
|
|
rectangle make_random_cropping_rect_resnet(
|
|
|
|
const matrix<rgb_pixel>& img,
|
|
|
|
dlib::rand& rnd
|
|
|
|
)
|
|
|
|
{
|
|
|
|
// figure out what rectangle we want to crop from the image
|
|
|
|
double mins = 0.466666666, maxs = 0.875;
|
|
|
|
auto scale = mins + rnd.get_random_double()*(maxs-mins);
|
|
|
|
auto size = scale*std::min(img.nr(), img.nc());
|
|
|
|
rectangle rect(size, size);
|
|
|
|
// randomly shift the box around
|
|
|
|
point offset(rnd.get_random_32bit_number()%(img.nc()-rect.width()),
|
|
|
|
rnd.get_random_32bit_number()%(img.nr()-rect.height()));
|
|
|
|
return move_rect(rect, offset);
|
|
|
|
}
|
|
|
|
|
|
|
|
// ----------------------------------------------------------------------------------------
|
|
|
|
|
|
|
|
void randomly_crop_images (
|
|
|
|
const matrix<rgb_pixel>& img,
|
|
|
|
dlib::array<matrix<rgb_pixel>>& crops,
|
|
|
|
dlib::rand& rnd,
|
|
|
|
long num_crops
|
|
|
|
)
|
|
|
|
{
|
|
|
|
std::vector<chip_details> dets;
|
|
|
|
for (long i = 0; i < num_crops; ++i)
|
|
|
|
{
|
|
|
|
auto rect = make_random_cropping_rect_resnet(img, rnd);
|
|
|
|
dets.push_back(chip_details(rect, chip_dims(227,227)));
|
|
|
|
}
|
|
|
|
|
|
|
|
extract_image_chips(img, dets, crops);
|
|
|
|
|
|
|
|
for (auto&& img : crops)
|
|
|
|
{
|
|
|
|
// Also randomly flip the image
|
|
|
|
if (rnd.get_random_double() > 0.5)
|
|
|
|
img = fliplr(img);
|
|
|
|
|
|
|
|
// And then randomly adjust the colors.
|
|
|
|
apply_random_color_offset(img, rnd);
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
// ----------------------------------------------------------------------------------------
|
|
|
|
|
|
|
|
int main(int argc, char** argv) try
|
|
|
|
{
|
2016-06-25 21:26:51 +08:00
|
|
|
if (argc == 1)
|
|
|
|
{
|
|
|
|
cout << "Give this program image files as command line arguments.\n" << endl;
|
2016-06-25 21:47:36 +08:00
|
|
|
cout << "You will also need a copy of the file resnet34_1000_imagenet_classifier.dnn " << endl;
|
|
|
|
cout << "available at http://dlib.net/files/resnet34_1000_imagenet_classifier.dnn.bz2" << endl;
|
2016-06-25 21:26:51 +08:00
|
|
|
cout << endl;
|
|
|
|
return 1;
|
|
|
|
}
|
|
|
|
|
2016-06-23 10:26:48 +08:00
|
|
|
std::vector<string> labels;
|
|
|
|
anet_type net;
|
|
|
|
deserialize("resnet34_1000_imagenet_classifier.dnn") >> net >> labels;
|
|
|
|
|
2016-06-26 06:31:21 +08:00
|
|
|
// Make a network with softmax as the final layer. We don't have to do this
|
|
|
|
// if we just want to output the single best prediction, since the anet_type
|
|
|
|
// already does this. But if we instead want to get the probability of each
|
|
|
|
// class as output we need to replace the last layer of the network with a
|
|
|
|
// softmax layer, which we do as follows:
|
2016-06-23 10:26:48 +08:00
|
|
|
softmax<anet_type::subnet_type> snet;
|
|
|
|
snet.subnet() = net.subnet();
|
|
|
|
|
|
|
|
dlib::array<matrix<rgb_pixel>> images;
|
|
|
|
matrix<rgb_pixel> img, crop;
|
|
|
|
|
|
|
|
dlib::rand rnd;
|
|
|
|
image_window win;
|
|
|
|
|
2016-06-26 06:31:21 +08:00
|
|
|
// Read images from the command prompt and print the top 5 best labels for each.
|
2016-06-23 10:26:48 +08:00
|
|
|
for (int i = 1; i < argc; ++i)
|
|
|
|
{
|
|
|
|
load_image(img, argv[i]);
|
|
|
|
const int num_crops = 16;
|
2016-06-25 21:26:51 +08:00
|
|
|
// Grab 16 random crops from the image. We will run all of them through the
|
|
|
|
// network and average the results.
|
2016-06-23 10:26:48 +08:00
|
|
|
randomly_crop_images(img, images, rnd, num_crops);
|
2016-06-25 21:26:51 +08:00
|
|
|
// p(i) == the probability the image contains object of class i.
|
2016-06-23 10:26:48 +08:00
|
|
|
matrix<float,1,1000> p = sum_rows(mat(snet(images.begin(), images.end())))/num_crops;
|
|
|
|
|
|
|
|
win.set_image(img);
|
2016-06-26 06:31:21 +08:00
|
|
|
// Print the 5 most probable labels
|
2016-06-23 10:26:48 +08:00
|
|
|
for (int k = 0; k < 5; ++k)
|
|
|
|
{
|
|
|
|
unsigned long predicted_label = index_of_max(p);
|
|
|
|
cout << p(predicted_label) << ": " << labels[predicted_label] << endl;
|
|
|
|
p(predicted_label) = 0;
|
|
|
|
}
|
|
|
|
|
2016-06-25 21:26:51 +08:00
|
|
|
cout << "Hit enter to process the next image";
|
2016-06-23 10:26:48 +08:00
|
|
|
cin.get();
|
|
|
|
}
|
|
|
|
|
|
|
|
}
|
|
|
|
catch(std::exception& e)
|
|
|
|
{
|
|
|
|
cout << e.what() << endl;
|
|
|
|
}
|
|
|
|
|