void
forward_network()
{
all_filters_update ();
all_dendrites_update ();
all_neurons_update ();
all_outputs_update ();
}
// Trains the visual search module
void
visual_search_thin_train(void)
{
set_neuron_layer_band_gaussian(&nl_v1_activation_map,HIGHEST_OUTPUT,(float)BAND_WIDTH/2.0,BAND_WIDTH);
filter_update(get_filter_by_name((char*)"nl_v1_activation_map_f_filter"));
train_neuron_layer((char*)"nl_v1_activation_map");
save_gaussian_filtered_training_pattern(&nl_v1_activation_map);
all_filters_update();
all_outputs_update();
}
void segment_draw_line (NEURON_LAYER *nl_segment, int x_center, int y_center, double alpha)
{
NEURON *neuron_vector = nl_segment->neuron_vector;
int width = nl_segment->dimentions.x;
int height = nl_segment->dimentions.y;
double dx = cos (alpha), dy = sin (alpha), i = 0;
int x0 = x_center - (int) (dx * 0.5 * (double) height);
int y0 = y_center - (int) (dy * 0.5 * (double) height);
for (i = 0.0; i < height; i += 1.0)
{
int x = x0 + (int) (dx * i);
int y = y0 + (int) (dy * i);
if (x < 0 || x >= width || y < 0 || y >= height)
break;
neuron_vector[y * width + x].output.ival = 0;
}
all_outputs_update ();
}
void
input_generator (INPUT_DESC *input, int status)
{
if (input->win == 0)
{
init_visual_search_thin (input);
}
else
{
if (status == MOVE)
{
check_input_bounds (input, input->wxd, input->wxd);
glutSetWindow (input->win);
input_display ();
update_input_filters();
all_dendrites_update ();
all_neurons_update ();
filter_update(get_filter_by_name((char*)"nl_v1_activation_map_f_filter")); //this filter is obligatory
all_outputs_update ();
}
}
}
void segment_set_neural_layer (NEURON_LAYER *nl_segment, NEURON_LAYER *nl_image, int center_x, int center_y)
{
int width = nl_segment->dimentions.x;
int height = nl_segment->dimentions.y;
char *reg_int = region_of_interest (nl_image, center_x, center_y, width, height);
char *img = segment_image (0.5, 50, 20, reg_int, nl_segment->dimentions.x, nl_segment->dimentions.y);
int x, y;
for (y = 0; y < height; y++)
{
for (x = 0; x < width; x++)
{
nl_segment->neuron_vector[y * nl_segment->dimentions.x + x].output.ival =
PIXEL(img[3 * (y * nl_segment->dimentions.x + x) + 0],
img[3 * (y * nl_segment->dimentions.x + x) + 1],
img[3 * (y * nl_segment->dimentions.x + x) + 2]);
}
}
all_outputs_update ();
free(img);
free(reg_int);
}
void input_generator (INPUT_DESC *input, int status)
{
if ((input->win != 0) && (status == MOVE))
{
update_input_neurons (input);
check_input_bounds (input, input->wxd, input->wyd);
glutSetWindow(input->win);
input_display ();
all_filters_update ();
all_outputs_update ();
}
if (input->win == 0)
{
int x, y;
make_input_image (input);
init (input);
glutInitWindowSize (input->ww, input->wh);
if (read_window_position (input->name, &x, &y))
glutInitWindowPosition (x, y);
else
glutInitWindowPosition (-1, -1);
input->win = glutCreateWindow (input->name);
glGenTextures (1, (GLuint *)(&(input->tex)));
input_init (input);
glutReshapeFunc (input_reshape);
glutDisplayFunc (input_display);
glutKeyboardFunc (keyboard);
glutPassiveMotionFunc (input_passive_motion);
glutMouseFunc (input_mouse);
}
}
void get_right_trunk_limit (NEURON_LAYER *nl_segment, int initial_x, int initial_y, int *x_limit, int *y_limit)
{
int xo, yo, i;
int trunk_value, value;
trunk_value = nl_segment->neuron_vector[initial_y * nl_segment->dimentions.x + initial_x].output.ival;
xo = initial_x;
yo = initial_y;
while(1)
{
nl_segment->neuron_vector[yo * nl_segment->dimentions.x + xo].output.ival = 0;
all_outputs_update ();
if ((xo + 1) < nl_segment->dimentions.x)
{
value = nl_segment->neuron_vector[yo * nl_segment->dimentions.x + (xo+1)].output.ival;
if (value == trunk_value)
{
xo++;
continue;
}
}
if ((xo + 1) < nl_segment->dimentions.x && (yo + 1) < nl_segment->dimentions.y)
{
value = nl_segment->neuron_vector[(yo+1) * nl_segment->dimentions.x + (xo+1)].output.ival;
if (value == trunk_value)
{
xo++;
yo++;
continue;
}
}
if ((xo + 1) < nl_segment->dimentions.x && (yo - 1) < nl_segment->dimentions.y)
{
value = nl_segment->neuron_vector[(yo-1) * nl_segment->dimentions.x + (xo+1)].output.ival;
if (value == trunk_value)
{
xo++;
yo--;
continue;
}
}
if ((xo + 1) < nl_segment->dimentions.x)
{
for(i = 0; i < nl_segment->dimentions.y; i++)
{
value = nl_segment->neuron_vector[i * nl_segment->dimentions.x + (xo+1)].output.ival;
if (value == trunk_value)
{
xo++;
yo = i;
break;
}
}
if (i < nl_segment->dimentions.y)
continue;
}
*x_limit = xo;
*y_limit = yo;
break;
}
}
