darknet/src/network.c

208 lines
6.1 KiB
C
Raw Normal View History

2013-11-13 22:50:38 +04:00
#include <stdio.h>
2013-11-04 23:11:01 +04:00
#include "network.h"
#include "image.h"
2013-11-13 22:50:38 +04:00
#include "data.h"
2013-11-04 23:11:01 +04:00
#include "connected_layer.h"
#include "convolutional_layer.h"
#include "maxpool_layer.h"
network make_network(int n)
{
network net;
net.n = n;
net.layers = calloc(net.n, sizeof(void *));
net.types = calloc(net.n, sizeof(LAYER_TYPE));
return net;
}
2013-11-13 22:50:38 +04:00
void forward_network(network net, double *input)
2013-11-04 23:11:01 +04:00
{
int i;
for(i = 0; i < net.n; ++i){
if(net.types[i] == CONVOLUTIONAL){
convolutional_layer layer = *(convolutional_layer *)net.layers[i];
2013-11-13 22:50:38 +04:00
forward_convolutional_layer(layer, input);
2013-11-04 23:11:01 +04:00
input = layer.output;
}
else if(net.types[i] == CONNECTED){
connected_layer layer = *(connected_layer *)net.layers[i];
2013-11-13 22:50:38 +04:00
forward_connected_layer(layer, input);
input = layer.output;
2013-11-04 23:11:01 +04:00
}
else if(net.types[i] == MAXPOOL){
maxpool_layer layer = *(maxpool_layer *)net.layers[i];
2013-11-13 22:50:38 +04:00
forward_maxpool_layer(layer, input);
2013-11-04 23:11:01 +04:00
input = layer.output;
}
}
}
void update_network(network net, double step)
{
int i;
for(i = 0; i < net.n; ++i){
if(net.types[i] == CONVOLUTIONAL){
convolutional_layer layer = *(convolutional_layer *)net.layers[i];
update_convolutional_layer(layer, step);
}
else if(net.types[i] == MAXPOOL){
//maxpool_layer layer = *(maxpool_layer *)net.layers[i];
}
else if(net.types[i] == CONNECTED){
connected_layer layer = *(connected_layer *)net.layers[i];
2013-11-13 22:50:38 +04:00
update_connected_layer(layer, step, .3, 0);
}
}
}
2013-11-13 22:50:38 +04:00
double *get_network_output_layer(network net, int i)
{
if(net.types[i] == CONVOLUTIONAL){
convolutional_layer layer = *(convolutional_layer *)net.layers[i];
return layer.output;
} else if(net.types[i] == MAXPOOL){
maxpool_layer layer = *(maxpool_layer *)net.layers[i];
return layer.output;
} else if(net.types[i] == CONNECTED){
connected_layer layer = *(connected_layer *)net.layers[i];
return layer.output;
}
return 0;
}
double *get_network_output(network net)
{
return get_network_output_layer(net, net.n-1);
}
double *get_network_delta_layer(network net, int i)
{
if(net.types[i] == CONVOLUTIONAL){
convolutional_layer layer = *(convolutional_layer *)net.layers[i];
return layer.delta;
} else if(net.types[i] == MAXPOOL){
maxpool_layer layer = *(maxpool_layer *)net.layers[i];
return layer.delta;
} else if(net.types[i] == CONNECTED){
connected_layer layer = *(connected_layer *)net.layers[i];
return layer.delta;
}
return 0;
}
double *get_network_delta(network net)
{
return get_network_delta_layer(net, net.n-1);
}
void learn_network(network net, double *input)
{
int i;
2013-11-13 22:50:38 +04:00
double *prev_input;
double *prev_delta;
for(i = net.n-1; i >= 0; --i){
if(i == 0){
2013-11-13 22:50:38 +04:00
prev_input = input;
prev_delta = 0;
}else{
prev_input = get_network_output_layer(net, i-1);
prev_delta = get_network_delta_layer(net, i-1);
}
if(net.types[i] == CONVOLUTIONAL){
convolutional_layer layer = *(convolutional_layer *)net.layers[i];
2013-11-13 22:50:38 +04:00
learn_convolutional_layer(layer, prev_input);
if(i != 0) backward_convolutional_layer(layer, prev_input, prev_delta);
}
else if(net.types[i] == MAXPOOL){
//maxpool_layer layer = *(maxpool_layer *)net.layers[i];
}
else if(net.types[i] == CONNECTED){
connected_layer layer = *(connected_layer *)net.layers[i];
2013-11-13 22:50:38 +04:00
learn_connected_layer(layer, prev_input);
if(i != 0) backward_connected_layer(layer, prev_input, prev_delta);
}
}
}
2013-11-13 22:50:38 +04:00
void train_network_batch(network net, batch b)
{
2013-11-13 22:50:38 +04:00
int i,j;
int k = get_network_output_size(net);
int correct = 0;
for(i = 0; i < b.n; ++i){
forward_network(net, b.images[i].data);
image o = get_network_image(net);
double *output = get_network_output(net);
double *delta = get_network_delta(net);
for(j = 0; j < k; ++j){
//printf("%f %f\n", b.truth[i][j], output[j]);
delta[j] = b.truth[i][j]-output[j];
if(fabs(delta[j]) < .5) ++correct;
//printf("%f\n", output[j]);
}
learn_network(net, b.images[i].data);
update_network(net, .00001);
}
2013-11-13 22:50:38 +04:00
printf("Accuracy: %f\n", (double)correct/b.n);
}
int get_network_output_size_layer(network net, int i)
{
if(net.types[i] == CONVOLUTIONAL){
convolutional_layer layer = *(convolutional_layer *)net.layers[i];
2013-11-13 22:50:38 +04:00
image output = get_convolutional_image(layer);
return output.h*output.w*output.c;
}
else if(net.types[i] == MAXPOOL){
maxpool_layer layer = *(maxpool_layer *)net.layers[i];
2013-11-13 22:50:38 +04:00
image output = get_maxpool_image(layer);
return output.h*output.w*output.c;
}
else if(net.types[i] == CONNECTED){
connected_layer layer = *(connected_layer *)net.layers[i];
return layer.outputs;
}
return 0;
}
2013-11-13 22:50:38 +04:00
int get_network_output_size(network net)
{
int i = net.n-1;
2013-11-13 22:50:38 +04:00
return get_network_output_size_layer(net, i);
}
image get_network_image_layer(network net, int i)
{
if(net.types[i] == CONVOLUTIONAL){
convolutional_layer layer = *(convolutional_layer *)net.layers[i];
2013-11-13 22:50:38 +04:00
return get_convolutional_image(layer);
}
else if(net.types[i] == MAXPOOL){
maxpool_layer layer = *(maxpool_layer *)net.layers[i];
2013-11-13 22:50:38 +04:00
return get_maxpool_image(layer);
}
return make_image(0,0,0);
}
2013-11-04 23:11:01 +04:00
image get_network_image(network net)
{
int i;
for(i = net.n-1; i >= 0; --i){
2013-11-13 22:50:38 +04:00
image m = get_network_image_layer(net, i);
if(m.h != 0) return m;
}
return make_image(1,1,1);
}
void visualize_network(network net)
{
int i;
for(i = 0; i < 1; ++i){
2013-11-04 23:11:01 +04:00
if(net.types[i] == CONVOLUTIONAL){
convolutional_layer layer = *(convolutional_layer *)net.layers[i];
2013-11-13 22:50:38 +04:00
visualize_convolutional_layer(layer);
2013-11-04 23:11:01 +04:00
}
2013-11-13 22:50:38 +04:00
}
2013-11-04 23:11:01 +04:00
}