def preprocess(self, x):
if self.type == "file":
return processing_utils.decode_base64_to_file(x)
elif self.type == "bytes":
return processing_utils.decode_base64_to_binary(x)
else:
raise ValueError("Unknown type: " + self.type + ". Please choose from: 'file', 'bytes'.")
def preprocess(self, x):
"""
By default, no pre-processing is applied to a microphone input file
"""
file_obj = processing_utils.decode_base64_to_file(x)
if self.preprocessing == "mfcc":
return processing_utils.generate_mfcc_features_from_audio_file(file_obj.name)
_, signal = scipy.io.wavfile.read(file_obj.name)
return signal
def preprocess(self, x):
name, data, is_local_example = x["name"], x["data"], x["is_local_example"]
if self.type == "file":
if is_local_example:
return open(name)
else:
return processing_utils.decode_base64_to_file(data)
elif self.type == "bytes":
if is_local_example:
with open(name, "rb") as file_data:
return file_data.read()
return processing_utils.decode_base64_to_binary(data)
else:
raise ValueError("Unknown type: " + str(self.type) + ". Please choose from: 'file', 'bytes'.")
def save_flagged_file(self, dir, label, data):
file = processing_utils.decode_base64_to_file(data)
old_file_name = file.name
output_dir = os.path.join(dir, label)
if os.path.exists(output_dir):
file_index = len(os.listdir(output_dir))
else:
os.mkdir(output_dir)
file_index = 0
new_file_name = str(file_index)
if "." in old_file_name:
uploaded_format = old_file_name.split(".")[-1].lower()
new_file_name += "." + uploaded_format
shutil.move(old_file_name, os.path.join(dir, label, new_file_name))
return label + "/" + new_file_name
def get_interpretation_neighbors(self, x):
file_obj = processing_utils.decode_base64_to_file(x)
x = scipy.io.wavfile.read(file_obj.name)
sample_rate, data = x
leave_one_out_sets = []
duration = data.shape[0]
boundaries = np.linspace(0, duration, self.interpretation_segments + 1).tolist()
boundaries = [round(boundary) for boundary in boundaries]
for index in range(len(boundaries) - 1):
leave_one_out_data = np.copy(data)
start, stop = boundaries[index], boundaries[index + 1]
leave_one_out_data[start:stop] = 0
file = tempfile.NamedTemporaryFile()
scipy.io.wavfile.write(file, sample_rate, leave_one_out_data)
out_data = processing_utils.encode_file_to_base64(file.name, type="audio", ext="wav")
leave_one_out_sets.append(out_data)
return leave_one_out_sets, {}, True
def save_flagged_file(self, dir, label, data, encryption_key):
if data is None:
return None
file = processing_utils.decode_base64_to_file(data, encryption_key)
label = "".join(
[char for char in label if char.isalnum() or char in "._- "])
old_file_name = file.name
output_dir = os.path.join(dir, label)
if os.path.exists(output_dir):
file_index = len(os.listdir(output_dir))
else:
os.makedirs(output_dir)
file_index = 0
new_file_name = str(file_index)
if "." in old_file_name:
uploaded_format = old_file_name.split(".")[-1].lower()
new_file_name += "." + uploaded_format
file.close()
shutil.move(old_file_name, os.path.join(dir, label, new_file_name))
return label + "/" + new_file_name
def deserialize(self, x):
return processing_utils.decode_base64_to_file(x).name
