def bot():
response = {"success": False}
if flask.request.method == "POST":
data = flask.request.get_json()
if "path" not in data:
return "missing path"
print(data)
try:
sound_file_path = download_remote_file(data["path"])
prediction = get_prediction(sound_file_path)
# indicate that the request was a success
response["success"] = True
response["predictions"] = prediction
print(response)
except Exception as e:
return flask.jsonify({"error": e})
# upload prediction to cloud storage
storage = StorageFactory.cloud()
storage.upload_prediction(source=sound_file_path,
model_type=MODEL_TYPE,
model_num=MODEL_NUM,
status=response["predictions"])
return flask.jsonify(response)
def predict():
# initialize the data dictionary that will be returned from the
# view
response = {"success": False}
# ensure an image was properly uploaded to our endpoint
if flask.request.method == "POST":
# make sure we received a file in the request
if not flask.request.files.get("file"):
response["error"] = "missing file key: 'file'"
return flask.jsonify(response)
# read the sound file
status, sound_file = save_file_from_request(flask.request)
prediction = get_prediction(sound_file)
response["predictions"] = prediction
# indicate that the request was a success
response["success"] = True
# upload prediction to cloud storage
storage = StorageFactory.cloud()
storage.upload_prediction(source=sound_file,
model_type=MODEL_TYPE,
model_num=MODEL_NUM,
status=response["predictions"])
# return the data dictionary as a JSON response
return flask.jsonify(response)
def in_cloud_cache(file_path):
storage = StorageFactory.cloud()
# get the file from the storage if possible
if storage.exists(file_path):
storage.save_file()
return file_path
def cloud_upload(self, model_type_path, exp_name, exp_id):
"""
Uploads the local models folder to the cloud
:param model_type_path:
:param exp_name:
:param exp_id:
:return:
"""
storage = StorageFactory.cloud()
storage.save_models_folder(model_type_path, exp_name, exp_id)
def load(from_cloud=True):
# The current served model based on the experiment type
global model
storage = StorageFactory.default()
file_path = storage.load_model(MODEL_TYPE, MODEL_NUM)
model = load_model(file_path)
# BUG fix - initializing the model with an empty vector
model.predict(np.zeros((1, 13, 30, 1)))
def download(self, url, output_file, upload=True):
# create the output file path
sound_file_path = os.path.join(self.destination_folder,
"{}.wav".format(output_file))
if self.debug:
print('downloading {} to {}'.format(output_file, sound_file_path))
(filename, headers) = urllib.request.urlretrieve(url)
sound = AudioSegment.from_mp3(filename)
# saving the file in wav format converted from mp3
sound.export(sound_file_path, format="wav")
# upload to cloud storage
if upload:
storage = StorageFactory.cloud()
storage.upload_wav(sound_file_path)
def save_file_from_request(request):
"""
Get a file object in the request and save it locally for predicition
:param request:
:return:
"""
file = request.files.get('file')
if not file:
full_path = None
file_uploaded = False
return file_uploaded, full_path
storage = StorageFactory.local()
full_path = storage.save_file(file, root=app.config['UPLOAD_FOLDER'])
file_uploaded = True
return file_uploaded, full_path
def test_connect_to_cloud_storage():
storage = StorageFactory.cloud()
print(storage)