def get_sample_lists(path):
t_images = load_dataset_paths(path, file_type='jpg')
const.dprint('Loading categories...\t')
training_list = []
training_labels = []
label_ids = []
for category in list(t_images):
if (const.USE_FULL_DATASET or (category in const.DATASET_CATEGORIES)) and (category != 'dataset_count'):
training_list.extend([os.path.join(path, category, i) for i in training_images[category][0:const.NUMBER_OF_TRAINING_IMAGES]])
training_labels.extend([len(label_ids)]*const.NUMBER_OF_TRAINING_IMAGES)
label_ids.append(category)
const.dprint('Done.\n')
return (training_list, training_labels, label_ids)
def get_sample_lists(path):
t_images = load_dataset_paths(path, file_type='jpg')
const.dprint('Loading categories...\t')
training_list = []
training_labels = []
label_ids = []
for category in list(t_images):
if (const.USE_FULL_DATASET or
(category
in const.DATASET_CATEGORIES)) and (category != 'dataset_count'):
training_list.extend([
os.path.join(path, category, i) for i in
training_images[category][0:const.NUMBER_OF_TRAINING_IMAGES]
])
training_labels.extend([len(label_ids)] *
const.NUMBER_OF_TRAINING_IMAGES)
label_ids.append(category)
const.dprint('Done.\n')
return (training_list, training_labels, label_ids)
for ii in training_images[category]:
p = os.path.join(path, category, ii)
p_data = os.path.join(data_path, category, ii +'.mat')
#try:
mat = image_lib.convert_to_matrix(p)
#except:
# print(p)
# print(category)
# input('pause')
with open(p_data, 'w') as f:
for aa in range(0, len(mat)):
for bb in range(0, len(mat[aa])):
f.write("{0}\n".format(mat[aa][bb]))
count += 1
console_lib.update_progress_bar(count/total_count, 'Resizing ' + category + ' ' + ii + ' ')
def create_neural_network(n_inputs, n_hidden, n_outputs, activation_function, learning_rate):
return neural_network.neural_network(n_inputs, n_hidden, n_outputs, activation_function, learning_rate)
def run_training(training_list, training_labels, neu_net):
for ii in range(0, len(training_list)):
matrix = image_lib.convert_to_matrix(training_list[ii])
neu_net.add_sample(matrix, const.RESIZE_IMAGE_DIMENSIONS, True, correct_label=training_labels[ii])
if const.PERFORM_PREPROCESSING:
preprocess_dataset(const.DATASET_PATH, resize_images=const.RESIZE_IMAGES, resize_image_dimensions=const.RESIZE_IMAGE_DIMENSIONS, store_data=const.CREATE_DATA_FILES, data_path=const.DATA_FILE_LOCATIONS)
const.dprint('Script Finished.\n')
f.write("{0}\n".format(mat[aa][bb]))
count += 1
console_lib.update_progress_bar(
count / total_count, 'Resizing ' + category + ' ' +
ii + ' ')
def create_neural_network(n_inputs, n_hidden, n_outputs, activation_function,
learning_rate):
return neural_network.neural_network(n_inputs, n_hidden, n_outputs,
activation_function, learning_rate)
def run_training(training_list, training_labels, neu_net):
for ii in range(0, len(training_list)):
matrix = image_lib.convert_to_matrix(training_list[ii])
neu_net.add_sample(matrix,
const.RESIZE_IMAGE_DIMENSIONS,
True,
correct_label=training_labels[ii])
if const.PERFORM_PREPROCESSING:
preprocess_dataset(const.DATASET_PATH,
resize_images=const.RESIZE_IMAGES,
resize_image_dimensions=const.RESIZE_IMAGE_DIMENSIONS,
store_data=const.CREATE_DATA_FILES,
data_path=const.DATA_FILE_LOCATIONS)
const.dprint('Script Finished.\n')