def get_model_data(self, argument, models):
# check argument
if argument not in models:
return None
# get model
model_json_config = models[argument]['json_config']
# load json
self.config.load_json(model_json_config, True)
# get model path
model_path = models[argument]['model_path']
model_name = os.path.basename(model_path)
model_size = get_formatted_file_size(model_path) if os.path.isfile(
model_path) else '121.12 MB'
model_classes = len(self.config.get_environment('classes'))
model_learning_epochs = self.config.getml('epochs')
model_date = get_changed_date(model_path) if os.path.isfile(
model_path) else '2019-10-20T11:54:25.125386+00:00'
model_version = '1.02'
return {
'model_name': model_name,
'model_size': model_size,
'model_classes': model_classes,
'model_learning_epochs': model_learning_epochs,
'model_date': model_date,
'model_version': model_version
}
def do_GET_prediction(self, argument, hook_results):
""" route GET /prediction """
model_data = None
hook_name = 'GET_prediction_get_model'
if hook_name in hook_results:
model_data = hook_results[hook_name]
if model_data is None:
model_data = self.get_empty_model_data()
database_path = get_database_path(argument)
database = get_database(argument)
used_model = self.get_template('used_model') % {
'MODEL_NAME': model_data['model_name'],
'MODEL_SIZE': model_data['model_size'],
'CLASSES': model_data['model_classes'],
'LEARNING_EPOCHS': model_data['model_learning_epochs'],
'MODEL_DATE': model_data['model_date'],
'VERSION': model_data['model_version'],
'DATABASE': os.path.basename(database_path),
'DATABASE_SIZE': get_formatted_file_size(database_path),
'DATABASE_VERSION': database['version']
}
# unknown model type
if argument not in self.allowed_model_types:
return False
# build html
html_form = self.get_template('form') % {
'ERROR_MESSAGE': '',
'TEXT_UPLOAD': self.TEXT_UPLOAD,
'MODEL_TYPE': argument
}
html_content = self.get_template(argument) % {
'PREDICTION_FORM': html_form,
'USED_MODEL': used_model
}
# send html to browser
self.respond_html(
self.get_template('body') % {'CONTENT': html_content})
return True
def get_model_data(argument):
if argument not in SimpleHTTPRequestHandler.allowed_model_types:
raise AssertionError('Unknown model type "%s"' % argument)
model_path = 'C:/Users/bjoern/data/processed/flower_10/flower_10_1.inceptionv3.best.17-0.95.h5'
model_name = os.path.basename(model_path)
model_size = get_formatted_file_size(model_path) if os.path.isfile(
model_path) else '121.12 MB'
model_classes = 12
model_learning_epochs = 20
model_date = get_changed_date(model_path) if os.path.isfile(
model_path) else '2019-10-20T11:54:25.125386+00:00'
model_version = '1.02'
return {
'model_name': model_name,
'model_size': model_size,
'model_classes': model_classes,
'model_learning_epochs': model_learning_epochs,
'model_date': model_date,
'model_version': model_version
}
def do_POST_prediction(self, argument, hook_results):
upload_data = self.write_upload_file()
model_data = None
hook_name = 'POST_prediction_get_model'
if hook_name in hook_results:
model_data = hook_results[hook_name]
if model_data is None:
model_data = self.get_empty_model_data()
database_path = get_database_path(argument)
database = get_database(argument)
used_model = self.get_template('used_model') % {
'MODEL_NAME': model_data['model_name'],
'MODEL_SIZE': model_data['model_size'],
'CLASSES': model_data['model_classes'],
'LEARNING_EPOCHS': model_data['model_learning_epochs'],
'MODEL_DATE': model_data['model_date'],
'VERSION': model_data['model_version'],
'DATABASE': os.path.basename(database_path),
'DATABASE_SIZE': get_formatted_file_size(database_path),
'DATABASE_VERSION': database['version']
}
# check if an error occurred
if upload_data['error']:
html_error = self.get_template('error') % upload_data['message']
html_form = self.get_template('form') % {
'ERROR_MESSAGE': html_error,
'TEXT_UPLOAD': self.TEXT_UPLOAD,
'MODEL_TYPE': argument
}
html_content = self.get_template(argument) % {
'PREDICTION_FORM': html_form,
'USED_MODEL': used_model
}
html_body = self.get_template('body') % {'CONTENT': html_content}
self.respond_html(html_body)
return True
# call post hook
hook_name = 'POST_%s' % ('prediction')
evaluation_data = self.call_hook(hook_name, argument, upload_data)
# get data from post hook
evaluated_file_web_size = get_formatted_file_size(
evaluation_data['evaluated_file'])
evaluated_file_web = evaluation_data['evaluated_file_web']
graph_file_web = evaluation_data['graph_file_web']
prediction_class = evaluation_data['prediction_class']
prediction_accuracy = evaluation_data['prediction_accuracy']
upload_form = self.get_template('form') % {
'ERROR_MESSAGE': '',
'TEXT_UPLOAD': self.TEXT_UPLOAD,
'MODEL_TYPE': argument
}
prediction_time = evaluation_data['prediction_time']
prediction_overview_array = evaluation_data[
'prediction_overview_array']
icons = {'flower': '🌻', 'food': '🍔'}
html_content = self.get_template('prediction') % {
'ICON': icons[argument],
'MODEL_TYPE': argument,
'MODEL_TYPE_TITLE': argument.title(),
'EVALUATED_FILE_WEB_SIZE': evaluated_file_web_size,
'EVALUATED_FILE_WEB': evaluated_file_web,
'PREDICTION_CLASS': prediction_class.replace('_', ' ').title(),
'PREDICTION_ACCURACY': '%.2f' % prediction_accuracy,
'GRAPH_FILE_WEB': graph_file_web,
'UPLOAD_FORM': upload_form,
'PREDICTION_TIME': prediction_time,
'USED_MODEL': used_model,
'DATABASE': json.dumps(database, cls=NumpyEncoder),
'EVALUATION_DATA': json.dumps(evaluation_data, cls=NumpyEncoder)
}
self.respond_html(
self.get_template('body') % {'CONTENT': html_content})
return True