def train(self):
samples = [];
print("Digit images preprocessing...");
for index_digit in range(0, 10, 1):
list_file_digit_sample = IMAGE_DIGIT_SAMPLES.GET_LIST_IMAGE_SAMPLES(index_digit);
for file_name in list_file_digit_sample:
data = read_image(file_name);
image_pattern = rgb2gray(data);
for index_pixel in range(len(image_pattern)):
if (image_pattern[index_pixel] < 128):
image_pattern[index_pixel] = 1;
else:
image_pattern[index_pixel] = 0;
samples += [ image_pattern ];
print("SOM initialization...");
self.__network = som(2, 5, type_conn.grid_four, None, True);
print("SOM training...");
self.__network.train(samples, 300);
print("SOM is ready...");
def click_image_load(self):
self.__user_pattern = [0 for i in range(32 * 32)]
Canvas.delete(self.__widget, "all")
index_digit = int(math.floor(random.random() * 10))
list_file_digit_sample = IMAGE_DIGIT_SAMPLES.GET_LIST_IMAGE_SAMPLES(
index_digit)
index_image = int(
math.floor(random.random() * len(list_file_digit_sample)))
file_name = list_file_digit_sample[index_image]
data = read_image(file_name)
image_pattern = rgb2gray(data)
for y in range(32):
for x in range(32):
linear_index = y * 32 + x
if (image_pattern[linear_index] < 128):
self.__user_pattern[linear_index] = 1
self.__widget.create_rectangle(x * 10,
y * 10,
x * 10 + 10,
y * 10 + 10,
fill=self.__color,
width=0)
def process(self,
image_source,
collect_dynamic=False,
order_color=0.9995,
order_object=0.999):
"""!
@brief Performs image segmentation.
@param[in] image_source (string): Path to image file that should be processed.
@param[in] collect_dynamic (bool): If 'True' then whole dynamic of each layer of the network is collected.
@param[in] order_color (double): Local synchronization order for the first layer - coloring segmentation.
@param[in] order_object (double): Local synchronization order for the second layer - object segmentation.
@return (syncsegm_analyser) Analyser of segmentation results by the network.
"""
self.__order_color = order_color
self.__order_object = order_object
data = read_image(image_source)
color_analyser = self.__analyse_colors(data, collect_dynamic)
if self.__object_radius is None:
return syncsegm_analyser(color_analyser, None)
object_segment_analysers = self.__analyse_objects(
image_source, color_analyser, collect_dynamic)
return syncsegm_analyser(color_analyser, object_segment_analysers)
def template_segmentation_image(source, color_radius, color_neighbors, object_radius, object_neighbors, noise_size):
data = read_image(source);
dbscan_instance = dbscan(data, color_radius, color_neighbors, True);
print("Segmentation: '", source, "', Dimensions:", len(data[0]));
dbscan_instance.process();
clusters = dbscan_instance.get_clusters();
real_clusters = [cluster for cluster in clusters if len(cluster) > noise_size];
print("Draw allocated color segments (back mask representation)...");
draw_image_mask_segments(source, real_clusters);
print("Draw allocated color segments (color segment representation)...");
draw_image_color_segments(source, real_clusters);
if (object_radius is None):
return;
# continue analysis
pointer_image = Image.open(source);
image_size = pointer_image.size;
object_colored_clusters = [];
for cluster in clusters:
coordinates = [];
for index in cluster:
y = int(floor(index / image_size[0]));
x = index - y * image_size[0];
coordinates.append([x, y]);
# perform clustering analysis of the colored objects
if (len(coordinates) < noise_size):
continue;
dbscan_instance = dbscan(coordinates, object_radius, object_neighbors, True);
dbscan_instance.process();
object_clusters = dbscan_instance.get_clusters();
# decode it
real_description_clusters = [];
for object_cluster in object_clusters:
real_description = [];
for index_object in object_cluster:
real_description.append(cluster[index_object]);
real_description_clusters.append(real_description);
if (len(real_description) > noise_size):
object_colored_clusters.append(real_description);
print("Draw allocated object segments (back mask representation)...");
draw_image_mask_segments(source, object_colored_clusters);
print("Draw allocated object segments (color segment representation)...");
draw_image_color_segments(source, object_colored_clusters);
def template_segmentation_image(source, color_radius, color_neighbors, object_radius, object_neighbors, noise_size):
data = read_image(source);
dbscan_instance = dbscan(data, color_radius, color_neighbors, True);
print("Segmentation: '", source, "', Dimensions:", len(data[0]));
dbscan_instance.process();
clusters = dbscan_instance.get_clusters();
real_clusters = [cluster for cluster in clusters if len(cluster) > noise_size];
print("Draw allocated color segments (back mask representation)...");
draw_image_mask_segments(source, real_clusters);
print("Draw allocated color segments (color segment representation)...");
draw_image_color_segments(source, real_clusters);
if (object_radius is None):
return;
# continue analysis
pointer_image = Image.open(source);
image_size = pointer_image.size;
object_colored_clusters = [];
for cluster in clusters:
coordinates = [];
for index in cluster:
y = int(floor(index / image_size[0]));
x = index - y * image_size[0];
coordinates.append([x, y]);
# perform clustering analysis of the colored objects
if (len(coordinates) < noise_size):
continue;
dbscan_instance = dbscan(coordinates, object_radius, object_neighbors, True);
dbscan_instance.process();
object_clusters = dbscan_instance.get_clusters();
# decode it
real_description_clusters = [];
for object_cluster in object_clusters:
real_description = [];
for index_object in object_cluster:
real_description.append(cluster[index_object]);
real_description_clusters.append(real_description);
if (len(real_description) > noise_size):
object_colored_clusters.append(real_description);
print("Draw allocated object segments (back mask representation)...");
draw_image_mask_segments(source, object_colored_clusters);
print("Draw allocated object segments (color segment representation)...");
draw_image_color_segments(source, object_colored_clusters);
def template_segmentation_image(image_file, parameters, steps, time, ccore_flag = True):
image = read_image(image_file);
stimulus = rgb2gray(image);
for pixel_index in range(len(stimulus)):
if (stimulus[pixel_index] < 235): stimulus[pixel_index] = 1;
else: stimulus[pixel_index] = 0;
if (parameters is None):
parameters = legion_parameters();
net = legion_network(len(stimulus), parameters, conn_type.GRID_FOUR, ccore = ccore_flag);
output_dynamic = net.simulate(steps, time, stimulus);
ensembles = output_dynamic.allocate_sync_ensembles();
draw_image_mask_segments(image_file, ensembles);
# draw_dynamics(output_dynamic.time, output_dynamic.output, x_title = "Time", y_title = "x(t)", separate = ensembles);
# just for checking correctness of results - let's use classical algorithm
dbscan_instance = dbscan(image, 3, 4, True);
dbscan_instance.process();
trustable_clusters = dbscan_instance.get_clusters();
draw_dynamics(output_dynamic.time, output_dynamic.output, x_title = "Time", y_title = "x(t)", separate = trustable_clusters);
def train(self):
samples = [];
print("Digit images preprocessing...");
for index_digit in range(0, 10, 1):
list_file_digit_sample = IMAGE_DIGIT_SAMPLES.GET_LIST_IMAGE_SAMPLES(index_digit);
for file_name in list_file_digit_sample:
data = read_image(file_name);
image_pattern = rgb2gray(data);
for index_pixel in range(len(image_pattern)):
if (image_pattern[index_pixel] < 128):
image_pattern[index_pixel] = 1;
else:
image_pattern[index_pixel] = 0;
samples += [ image_pattern ];
print("SOM initialization...");
self.__network = som(2, 5, type_conn.grid_four, None, True);
print("SOM training...");
self.__network.train(samples, 300);
print("SOM is ready...");
def template_segmentation_image(image_file, parameters, steps, time, ccore_flag = True):
image = read_image(image_file);
stimulus = rgb2gray(image);
for pixel_index in range(len(stimulus)):
if (stimulus[pixel_index] < 235): stimulus[pixel_index] = 1;
else: stimulus[pixel_index] = 0;
if (parameters is None):
parameters = legion_parameters();
net = legion_network(len(stimulus), parameters, conn_type.GRID_FOUR, ccore = ccore_flag);
output_dynamic = net.simulate(steps, time, stimulus);
ensembles = output_dynamic.allocate_sync_ensembles();
draw_image_mask_segments(image_file, ensembles);
# draw_dynamics(output_dynamic.time, output_dynamic.output, x_title = "Time", y_title = "x(t)", separate = ensembles);
# just for checking correctness of results - let's use classical algorithm
dbscan_instance = dbscan(image, 3, 4, True);
dbscan_instance.process();
trustable_clusters = dbscan_instance.get_clusters();
draw_dynamics(output_dynamic.time, output_dynamic.output, x_title = "Time", y_title = "x(t)", separate = trustable_clusters);
def template_segmentation_image(source, start_centers):
data = read_image(source)
kmeans_instance = kmeans(data, start_centers)
kmeans_instance.process()
clusters = kmeans_instance.get_clusters()
draw_image_mask_segments(source, clusters)
def template_segmentation_image(source, start_centers):
data = read_image(source);
kmeans_instance = kmeans(data, start_centers);
kmeans_instance.process();
clusters = kmeans_instance.get_clusters();
draw_image_mask_segments(source, clusters);
def template_segmentation(source, levels, threshold):
data = read_image(source)
bang_instance = bang(data, levels, threshold)
bang_instance.process()
clusters = bang_instance.get_clusters()
draw_image_mask_segments(source, clusters)
def testDrawSegmentationResultNoFailure(self):
data = utils.read_image(IMAGE_SIMPLE_SAMPLES.IMAGE_SIMPLE01);
kmeans_instance = kmeans(data, [[255, 0, 0], [0, 0, 255], [180, 136, 0], [255, 255, 255]]);
kmeans_instance.process();
clusters = kmeans_instance.get_clusters();
utils.draw_image_mask_segments(IMAGE_SIMPLE_SAMPLES.IMAGE_SIMPLE01, clusters);
utils.draw_image_color_segments(IMAGE_SIMPLE_SAMPLES.IMAGE_SIMPLE01, clusters);
def template_segmentation(source, levels, threshold):
data = read_image(source)
bang_instance = bang(data, levels, threshold)
bang_instance.process()
clusters = bang_instance.get_clusters()
draw_image_mask_segments(source, clusters)
def template_segmentation_image(image, parameters, simulation_time, brightness, scale_color = True, fastlinking = False, show_spikes = False, ccore_flag = True):
image_source = Image.open(image);
image_size = image_source.size;
width = image_size[0];
height = image_size[1];
stimulus = read_image(image);
stimulus = rgb2gray(stimulus);
if (brightness != None):
for pixel_index in range(len(stimulus)):
if (stimulus[pixel_index] < brightness): stimulus[pixel_index] = 1;
else: stimulus[pixel_index] = 0;
else:
maximum_stimulus = float(max(stimulus));
minimum_stimulus = float(min(stimulus));
delta = maximum_stimulus - minimum_stimulus;
for pixel_index in range(len(stimulus)):
if (scale_color is True):
stimulus[pixel_index] = 1.0 - ((float(stimulus[pixel_index]) - minimum_stimulus) / delta);
else:
stimulus[pixel_index] = float(stimulus[pixel_index]) / 255;
if (parameters is None):
parameters = pcnn_parameters();
parameters.AF = 0.1;
parameters.AL = 0.1;
parameters.AT = 0.8;
parameters.VF = 1.0;
parameters.VL = 1.0;
parameters.VT = 30.0;
parameters.W = 1.0;
parameters.M = 1.0;
parameters.FAST_LINKING = fastlinking;
net = pcnn_network(len(stimulus), parameters, conn_type.GRID_EIGHT, height = height, width = width, ccore = ccore_flag);
output_dynamic = net.simulate(simulation_time, stimulus);
pcnn_visualizer.show_output_dynamic(output_dynamic);
ensembles = output_dynamic.allocate_sync_ensembles();
draw_image_mask_segments(image, ensembles);
pcnn_visualizer.show_time_signal(output_dynamic);
if (show_spikes is True):
spikes = output_dynamic.allocate_spike_ensembles();
draw_image_mask_segments(image, spikes);
pcnn_visualizer.animate_spike_ensembles(output_dynamic, image_size);
def template_segmentation_image(image, parameters, simulation_time, brightness, scale_color = True, fastlinking = False, show_spikes = False, ccore_flag = True):
image_source = Image.open(image);
image_size = image_source.size;
width = image_size[0];
height = image_size[1];
stimulus = read_image(image);
stimulus = rgb2gray(stimulus);
if (brightness != None):
for pixel_index in range(len(stimulus)):
if (stimulus[pixel_index] < brightness): stimulus[pixel_index] = 1;
else: stimulus[pixel_index] = 0;
else:
maximum_stimulus = float(max(stimulus));
minimum_stimulus = float(min(stimulus));
delta = maximum_stimulus - minimum_stimulus;
for pixel_index in range(len(stimulus)):
if (scale_color is True):
stimulus[pixel_index] = 1.0 - ((float(stimulus[pixel_index]) - minimum_stimulus) / delta);
else:
stimulus[pixel_index] = float(stimulus[pixel_index]) / 255;
if (parameters is None):
parameters = pcnn_parameters();
parameters.AF = 0.1;
parameters.AL = 0.1;
parameters.AT = 0.8;
parameters.VF = 1.0;
parameters.VL = 1.0;
parameters.VT = 30.0;
parameters.W = 1.0;
parameters.M = 1.0;
parameters.FAST_LINKING = fastlinking;
net = pcnn_network(len(stimulus), parameters, conn_type.GRID_EIGHT, height = height, width = width, ccore = ccore_flag);
output_dynamic = net.simulate(simulation_time, stimulus);
pcnn_visualizer.show_output_dynamic(output_dynamic);
ensembles = output_dynamic.allocate_sync_ensembles();
draw_image_mask_segments(image, ensembles);
pcnn_visualizer.show_time_signal(output_dynamic);
if (show_spikes is True):
spikes = output_dynamic.allocate_spike_ensembles();
draw_image_mask_segments(image, spikes);
pcnn_visualizer.animate_spike_ensembles(output_dynamic, image_size);
def template_segmentation_image_amount_colors(source, amount):
data = read_image(source)
centers = kmeans_plusplus_initializer(
data, amount,
kmeans_plusplus_initializer.FARTHEST_CENTER_CANDIDATE).initialize()
kmeans_instance = kmeans(data, centers)
kmeans_instance.process()
clusters = kmeans_instance.get_clusters()
draw_image_mask_segments(source, clusters)
def template_segmentation_image(source, map_som_size = [5, 5], radius = 128.0, sync_order = 0.998, show_dyn = False, show_som_map = False):
data = read_image(source);
network = syncsom(data, map_som_size[0], map_som_size[1], 1.0);
(ticks, (dyn_time, dyn_phase)) = timedcall(network.process, show_dyn, sync_order);
print("Sample: ", source, "\t\tExecution time: ", ticks, "\t\tWinners: ", network.som_layer.get_winner_number(), "\n");
if (show_dyn is True):
draw_dynamics(dyn_time, dyn_phase);
clusters = network.get_clusters();
draw_image_mask_segments(source, clusters);
def template_segmentation_image(source, map_som_size = [5, 5], average_neighbors = 5, sync_order = 0.998, show_dyn = False, show_som_map = False):
data = read_image(source);
network = syncsom(data, map_som_size[0], map_som_size[1]);
(ticks, (dyn_time, dyn_phase)) = timedcall(network.process, average_neighbors, show_dyn, sync_order);
print("Sample: ", source, "\t\tExecution time: ", ticks, "\t\tWinners: ", network.som_layer.get_winner_number(), "\n");
if (show_dyn is True):
draw_dynamics(dyn_time, dyn_phase);
clusters = network.get_clusters();
draw_image_mask_segments(source, clusters);
def click_image_load(self):
self.__user_pattern = [ 0 for i in range(32 * 32) ];
Canvas.delete(self.__widget, "all");
index_digit = int(math.floor(random.random() * 10));
list_file_digit_sample = IMAGE_DIGIT_SAMPLES.GET_LIST_IMAGE_SAMPLES(index_digit);
index_image = int(math.floor( random.random() * len(list_file_digit_sample) ));
file_name = list_file_digit_sample[index_image];
data = read_image(file_name);
image_pattern = rgb2gray(data);
for y in range(32):
for x in range(32):
linear_index = y * 32 + x;
if (image_pattern[linear_index] < 128):
self.__user_pattern[linear_index] = 1;
self.__widget.create_rectangle(x * 10, y * 10, x * 10 + 10, y * 10 + 10, fill = self.__color, width = 0);
def template_recognition_image(images, steps, time, corruption=0.1):
samples = []
for file_name in images:
data = read_image(file_name)
image_pattern = rgb2gray(data)
for index_pixel in range(len(image_pattern)):
if (image_pattern[index_pixel] < 128):
image_pattern[index_pixel] = 1
else:
image_pattern[index_pixel] = -1
samples += [image_pattern]
net = syncpr(len(samples[0]), 0.3, 0.3, ccore=True)
net.train(samples)
# Recognize the each learned pattern
for i in range(len(samples)):
sync_output_dynamic = net.simulate(steps, time, samples[i],
solve_type.RK4, True)
syncpr_visualizer.show_output_dynamic(sync_output_dynamic)
syncpr_visualizer.show_pattern(sync_output_dynamic, 10, 10)
# corrupt a little bit by black and white pixels
for _ in range(math.floor(len(samples[i]) * corruption)):
random.seed()
random_pixel = math.floor(random.random() * len(samples[i]))
samples[i][random_pixel] = 1
random_pixel = math.floor(random.random() * len(samples[i]))
samples[i][random_pixel] = -1
sync_output_dynamic = net.simulate(steps, time, samples[i],
solve_type.RK4, True)
syncpr_visualizer.show_output_dynamic(sync_output_dynamic)
syncpr_visualizer.show_pattern(sync_output_dynamic, 10, 10)
syncpr_visualizer.animate_pattern_recognition(
sync_output_dynamic, 10, 10, title="Pattern Recognition")
def template_recognition_image(images, steps, time, corruption = 0.1):
samples = [];
for file_name in images:
data = read_image(file_name);
image_pattern = rgb2gray(data);
for index_pixel in range(len(image_pattern)):
if (image_pattern[index_pixel] < 128):
image_pattern[index_pixel] = 1;
else:
image_pattern[index_pixel] = -1;
samples += [ image_pattern ];
net = syncpr(len(samples[0]), 0.3, 0.3, ccore = True);
net.train(samples);
# Recognize the each learned pattern
for i in range(len(samples)):
sync_output_dynamic = net.simulate(steps, time, samples[i], solve_type.RK4, True);
syncpr_visualizer.show_output_dynamic(sync_output_dynamic);
syncpr_visualizer.show_pattern(sync_output_dynamic, 10, 10);
# corrupt a little bit by black and white pixels
for _ in range( math.floor(len(samples[i]) * corruption) ):
random.seed();
random_pixel = math.floor(random.random() * len(samples[i]));
samples[i][random_pixel] = 1;
random_pixel = math.floor(random.random() * len(samples[i]));
samples[i][random_pixel] = -1;
sync_output_dynamic = net.simulate(steps, time, samples[i], solve_type.RK4, True);
syncpr_visualizer.show_output_dynamic(sync_output_dynamic);
syncpr_visualizer.show_pattern(sync_output_dynamic, 10, 10);
syncpr_visualizer.animate_pattern_recognition(sync_output_dynamic, 10, 10, title = "Pattern Recognition");
def segmentation_double_t():
image = read_image(IMAGE_SIMPLE_SAMPLES.IMAGE_SIMPLE10);
image = rgb2gray(image);
for pixel_index in range(len(image)):
if (image[pixel_index] < 128):
image[pixel_index] = 1;
else:
image[pixel_index] = 0;
params = pcnn_parameters();
params.AF = 0.1;
params.AL = 0.1;
params.AT = 0.8;
params.VF = 1.0;
params.VL = 1.0;
params.VT = 20.0;
params.W = 1.0;
params.M = 1.0;
ensembles = template_dynamic_pcnn(32 * 32, 28, image, params, conn_type.GRID_EIGHT, False);
draw_image_mask_segments(IMAGE_SIMPLE_SAMPLES.IMAGE_SIMPLE10, ensembles);
def segmentation_double_t():
image = read_image(IMAGE_SIMPLE_SAMPLES.IMAGE_SIMPLE10);
image = rgb2gray(image);
for pixel_index in range(len(image)):
if (image[pixel_index] < 128):
image[pixel_index] = 1;
else:
image[pixel_index] = 0;
params = pcnn_parameters();
params.AF = 0.1;
params.AL = 0.1;
params.AT = 0.8;
params.VF = 1.0;
params.VL = 1.0;
params.VT = 20.0;
params.W = 1.0;
params.M = 1.0;
ensembles = template_dynamic_pcnn(32 * 32, 28, image, params, conn_type.GRID_EIGHT, False);
draw_image_mask_segments(IMAGE_SIMPLE_SAMPLES.IMAGE_SIMPLE10, ensembles);
def process(self, image_source, collect_dynamic = False, order_color = 0.9995, order_object = 0.999):
"""!
@brief Performs image segmentation.
@param[in] image_source (string): Path to image file that should be processed.
@param[in] collect_dynamic (bool): If 'True' then whole dynamic of each layer of the network is collected.
@param[in] order_color (double): Local synchronization order for the first layer - coloring segmentation.
@param[in] order_object (double): Local synchronization order for the second layer - object segmentation.
@return (syncsegm_analyser) Analyser of segmentation results by the network.
"""
self.__color_radius = order_color;
self.__order_object = order_object;
data = read_image(image_source);
color_analyser = self.__analyse_colors(data, collect_dynamic);
if (self.__object_radius is None):
return syncsegm_analyser(color_analyser, None);
object_segment_analysers = self.__analyse_objects(image_source, color_analyser, collect_dynamic);
return syncsegm_analyser(color_analyser, object_segment_analysers);
def template_image_segmentation(image_file, steps, time, dynamic_file_prefix):
image = read_image(image_file)
stimulus = rgb2gray(image)
params = hhn_parameters()
params.deltah = 650
params.w1 = 0.1
params.w2 = 9.0
params.w3 = 5.0
params.threshold = -10
stimulus = [255.0 - pixel for pixel in stimulus]
divider = max(stimulus) / 50.0
stimulus = [int(pixel / divider) for pixel in stimulus]
t, dyn_peripheral, dyn_central = None, None, None
if (not os.path.exists(dynamic_file_prefix + 'dynamic_time.txt') or
not os.path.exists(dynamic_file_prefix + 'dynamic_peripheral.txt')
or not os.path.exists(dynamic_file_prefix +
'dynamic_dyn_central.txt')):
print(
"File with output dynamic is not found - simulation will be performed - it may take some time, be patient."
)
net = hhn_network(len(stimulus), stimulus, params, ccore=True)
(t, dyn_peripheral, dyn_central) = net.simulate(steps, time)
print("Store dynamic to save time for simulation next time.")
with open(dynamic_file_prefix + 'dynamic_time.txt',
'wb') as file_descriptor:
pickle.dump(t, file_descriptor)
with open(dynamic_file_prefix + 'dynamic_peripheral.txt',
'wb') as file_descriptor:
pickle.dump(dyn_peripheral, file_descriptor)
with open(dynamic_file_prefix + 'dynamic_dyn_central.txt',
'wb') as file_descriptor:
pickle.dump(dyn_central, file_descriptor)
else:
print("Load output dynamic from file.")
with open(dynamic_file_prefix + 'dynamic_time.txt',
'rb') as file_descriptor:
t = pickle.load(file_descriptor)
with open(dynamic_file_prefix + 'dynamic_peripheral.txt',
'rb') as file_descriptor:
dyn_peripheral = pickle.load(file_descriptor)
with open(dynamic_file_prefix + 'dynamic_dyn_central.txt',
'rb') as file_descriptor:
dyn_central = pickle.load(file_descriptor)
animate_segmentation(t, dyn_peripheral, image_file, 200)
# just for checking correctness of results - let's use classical algorithm
if (False):
dbscan_instance = dbscan(image, 3, 4, True)
dbscan_instance.process()
trustable_clusters = dbscan_instance.get_clusters()
amount_canvases = len(trustable_clusters) + 2
visualizer = dynamic_visualizer(amount_canvases,
x_title="Time",
y_title="V",
y_labels=False)
visualizer.append_dynamics(t, dyn_peripheral, 0, trustable_clusters)
visualizer.append_dynamics(t, dyn_central, amount_canvases - 2, True)
visualizer.show()
def template_image_segmentation(image_file, steps, time, dynamic_file_prefix):
image = read_image(image_file);
stimulus = rgb2gray(image);
params = hhn_parameters();
params.deltah = 650;
params.w1 = 0.1;
params.w2 = 9.0;
params.w3 = 5.0;
params.threshold = -10;
stimulus = [255.0 - pixel for pixel in stimulus];
divider = max(stimulus) / 50.0;
stimulus = [int(pixel / divider) for pixel in stimulus];
t, dyn_peripheral, dyn_central = None, None, None;
if ( not os.path.exists(dynamic_file_prefix + 'dynamic_time.txt') or
not os.path.exists(dynamic_file_prefix + 'dynamic_peripheral.txt') or
not os.path.exists(dynamic_file_prefix + 'dynamic_dyn_central.txt') ):
print("File with output dynamic is not found - simulation will be performed - it may take some time, be patient.");
net = hhn_network(len(stimulus), stimulus, params, ccore=True);
(t, dyn_peripheral, dyn_central) = net.simulate(steps, time);
print("Store dynamic to save time for simulation next time.");
with open(dynamic_file_prefix + 'dynamic_time.txt', 'wb') as file_descriptor:
pickle.dump(t, file_descriptor);
with open(dynamic_file_prefix + 'dynamic_peripheral.txt', 'wb') as file_descriptor:
pickle.dump(dyn_peripheral, file_descriptor);
with open(dynamic_file_prefix + 'dynamic_dyn_central.txt', 'wb') as file_descriptor:
pickle.dump(dyn_central, file_descriptor);
else:
print("Load output dynamic from file.");
with open (dynamic_file_prefix + 'dynamic_time.txt', 'rb') as file_descriptor:
t = pickle.load(file_descriptor);
with open (dynamic_file_prefix + 'dynamic_peripheral.txt', 'rb') as file_descriptor:
dyn_peripheral = pickle.load(file_descriptor);
with open (dynamic_file_prefix + 'dynamic_dyn_central.txt', 'rb') as file_descriptor:
dyn_central = pickle.load(file_descriptor);
animate_segmentation(t, dyn_peripheral, image_file, 200);
# just for checking correctness of results - let's use classical algorithm
if (False):
dbscan_instance = dbscan(image, 3, 4, True);
dbscan_instance.process();
trustable_clusters = dbscan_instance.get_clusters();
amount_canvases = len(trustable_clusters) + 2;
visualizer = dynamic_visualizer(amount_canvases, x_title = "Time", y_title = "V", y_labels = False);
visualizer.append_dynamics(t, dyn_peripheral, 0, trustable_clusters);
visualizer.append_dynamics(t, dyn_central, amount_canvases - 2, True);
visualizer.show();
def template_segmentation_image(source, color_radius, object_radius, noise_size, show_dyn):
data = read_image(source);
print("Pixel dimension: ", len(data[0]));
network = syncnet(data, color_radius, ccore = True);
print("Network has been created");
(ticks, analyser) = timedcall(network.process, 0.9995, solve_type.FAST, show_dyn);
print("Sample: ", source, "\t\tExecution time: ", ticks, "\n");
if (show_dyn is True):
sync_visualizer.show_output_dynamic(analyser);
clusters = analyser.allocate_clusters();
real_clusters = [cluster for cluster in clusters if len(cluster) > noise_size];
draw_image_mask_segments(source, real_clusters);
if (object_radius is None):
return;
# continue analysis
pointer_image = Image.open(source);
image_size = pointer_image.size;
object_colored_clusters = [];
object_colored_dynamics = [];
total_dyn = [];
for cluster in clusters:
coordinates = [];
for index in cluster:
y = floor(index / image_size[0]);
x = index - y * image_size[0];
coordinates.append([x, y]);
print(coordinates);
# perform clustering analysis of the colored objects
if (network is not None):
del network;
network = None;
if (len(coordinates) < noise_size):
continue;
network = syncnet(coordinates, object_radius, ccore = True);
analyser = network.process(0.999, solve_type.FAST, show_dyn);
if (show_dyn is True):
object_colored_dynamics.append( (analyser.time, analyser.output) );
object_clusters = analyser.allocate_clusters();
# decode it
real_description_clusters = [];
for object_cluster in object_clusters:
real_description = [];
for index_object in object_cluster:
real_description.append(cluster[index_object]);
real_description_clusters.append(real_description);
if (len(real_description) > noise_size):
object_colored_clusters.append(real_description);
# draw_image_mask_segments(source, [ cluster ]);
# draw_image_mask_segments(source, real_description_clusters);
draw_image_mask_segments(source, object_colored_clusters);
if (show_dyn is True):
draw_dynamics_set(object_colored_dynamics, None, None, None, [0, 2 * 3.14], False, False);
# an example of image color segmentation.
from pyclustering.utils import draw_image_mask_segments, read_image
from pyclustering.samples.definitions import IMAGE_SIMPLE_SAMPLES
from pyclustering.cluster.kmeans import kmeans
from pyclustering.cluster.center_initializer import kmeans_plusplus_initializer
# load image from the pyclustering collection.
data = read_image(IMAGE_SIMPLE_SAMPLES.IMAGE_SIMPLE_BEACH)
# set initial centers for K-Means algorithm.
amount_initial_centers = 3
initial_centers = kmeans_plusplus_initializer(
data, amount_initial_centers).initialize()
#start_centers = [ [153, 217, 234, 128], [0, 162, 232, 128], [34, 177, 76, 128], [255, 242, 0, 128] ]
# create K-Means algorithm instance.
kmeans_instance = kmeans(data, initial_centers)
# start processing.
kmeans_instance.process()
# obtain clusters that are considered as segments.
segments = kmeans_instance.get_clusters()
# show image segmentation results.
draw_image_mask_segments(IMAGE_SIMPLE_SAMPLES.IMAGE_SIMPLE_BEACH, segments)
# an example of image color segmentation.
from pyclustering.utils import draw_image_mask_segments, read_image
from pyclustering.samples.definitions import IMAGE_SIMPLE_SAMPLES
from pyclustering.cluster.kmeans import kmeans
from pyclustering.cluster.center_initializer import kmeans_plusplus_initializer
# load image from the pyclustering collection.
data = read_image("4_projeta/inner_img_rgb.tif")
# set initial centers for K-Means algorithm.
amount_initial_centers = 2
initial_centers = kmeans_plusplus_initializer(
data, amount_initial_centers).initialize()
# create K-Means algorithm instance.
kmeans_instance = kmeans(data, initial_centers)
# start processing.
kmeans_instance.process()
# obtain clusters that are considered as segments.
segments = kmeans_instance.get_clusters()
print(len(segments[1]))
# show image segmentation results.
clusterized = draw_image_mask_segments("4_projeta/inner_img_rgb.tif", segments)
clusterized.save('5_extrai_linhas/grupos.tif')
def template_segmentation_image(image,
parameters,
simulation_time,
brightness,
scale_color=True,
fastlinking=False,
show_spikes=False,
ccore_flag=True):
image_source = Image.open(image)
image_size = image_source.size
width = image_size[0]
height = image_size[1]
stimulus = read_image(image)
stimulus = rgb2gray(stimulus)
if brightness is not None:
for pixel_index in range(len(stimulus)):
if stimulus[pixel_index] < brightness:
stimulus[pixel_index] = 1
else:
stimulus[pixel_index] = 0
else:
maximum_stimulus = float(max(stimulus))
minimum_stimulus = float(min(stimulus))
delta = maximum_stimulus - minimum_stimulus
for pixel_index in range(len(stimulus)):
if scale_color is True:
stimulus[pixel_index] = 1.0 - (
(float(stimulus[pixel_index]) - minimum_stimulus) / delta)
else:
stimulus[pixel_index] = float(stimulus[pixel_index]) / 255
if parameters is None:
parameters = pcnn_parameters()
parameters.AF = 0.1
parameters.AL = 0.1
parameters.AT = 0.8
parameters.VF = 1.0
parameters.VL = 1.0
parameters.VT = 30.0
parameters.W = 1.0
parameters.M = 1.0
parameters.FAST_LINKING = fastlinking
net = pcnn_network(len(stimulus),
parameters,
conn_type.GRID_EIGHT,
height=height,
width=width,
ccore=ccore_flag)
output_dynamic = net.simulate(simulation_time, stimulus)
pcnn_visualizer.show_output_dynamic(output_dynamic)
ensembles = output_dynamic.allocate_sync_ensembles()
draw_image_mask_segments(image, ensembles)
pcnn_visualizer.show_time_signal(output_dynamic)
if show_spikes is True:
spikes = output_dynamic.allocate_spike_ensembles()
draw_image_mask_segments(image, spikes)
pcnn_visualizer.animate_spike_ensembles(output_dynamic, image_size)
def template_segmentation_image(source, color_radius, object_radius, noise_size, show_dyn):
data = read_image(source);
print("Pixel dimension: ", len(data[0]));
network = syncnet(data, color_radius, ccore = True);
print("Network has been created");
(ticks, analyser) = timedcall(network.process, 0.9995, solve_type.FAST, show_dyn);
print("Sample: ", source, "\t\tExecution time: ", ticks, "\n");
if (show_dyn is True):
sync_visualizer.show_output_dynamic(analyser);
clusters = analyser.allocate_clusters();
real_clusters = [cluster for cluster in clusters if len(cluster) > noise_size];
draw_image_mask_segments(source, real_clusters);
if (object_radius is None):
return;
# continue analysis
pointer_image = Image.open(source);
image_size = pointer_image.size;
object_colored_clusters = [];
object_colored_dynamics = [];
total_dyn = [];
for cluster in clusters:
coordinates = [];
for index in cluster:
y = floor(index / image_size[0]);
x = index - y * image_size[0];
coordinates.append([x, y]);
print(coordinates);
# perform clustering analysis of the colored objects
if (network is not None):
del network;
network = None;
if (len(coordinates) < noise_size):
continue;
network = syncnet(coordinates, object_radius, ccore = True);
analyser = network.process(0.999, solve_type.FAST, show_dyn);
if (show_dyn is True):
object_colored_dynamics.append( (analyser.time, analyser.output) );
object_clusters = analyser.allocate_clusters();
# decode it
real_description_clusters = [];
for object_cluster in object_clusters:
real_description = [];
for index_object in object_cluster:
real_description.append(cluster[index_object]);
real_description_clusters.append(real_description);
if (len(real_description) > noise_size):
object_colored_clusters.append(real_description);
# draw_image_mask_segments(source, [ cluster ]);
# draw_image_mask_segments(source, real_description_clusters);
draw_image_mask_segments(source, object_colored_clusters);
if (show_dyn is True):
draw_dynamics_set(object_colored_dynamics, None, None, None, [0, 2 * 3.14], False, False);
