def __init__(self, path_list, params):
self.path_list = path_list
self.params = params
# culculate the number of dimentions
dims = params["feature"]["n_mels"] * params["feature"]["n_frames"]
# iterate to file_to_vector_array()
for idx in tqdm(range(len(path_list))):
vectors = com.file_to_vectors(
path_list[idx],
n_mels=params["feature"]["n_mels"],
n_frames=params["feature"]["n_frames"],
n_fft=params["feature"]["n_fft"],
hop_length=params["feature"]["hop_length"],
power=params["feature"]["power"])
vectors = vectors[::params["feature"]["n_hop_frames"], :]
if idx == 0:
data = np.zeros((len(path_list) * vectors.shape[0], dims),
np.float32)
data[vectors.shape[0] * idx:vectors.shape[0] *
(idx + 1), :] = vectors
self.feat_data = data
def file_list_to_data(file_list,
msg="calc...",
n_mels=64,
n_frames=5,
n_hop_frames=1,
n_fft=1024,
hop_length=512,
power=2.0):
"""
convert the file_list to a vector array.
file_to_vector_array() is iterated, and the output vector array is concatenated.
file_list : list [ str ]
.wav filename list of dataset
msg : str ( default = "calc..." )
description for tqdm.
this parameter will be input into "desc" param at tqdm.
return : numpy.array( numpy.array( float ) )
data for training (this function is not used for test.)
* dataset.shape = (number of feature vectors, dimensions of feature vectors)
"""
# calculate the number of dimensions
dims = n_mels * n_frames
# iterate file_to_vector_array()
for idx in tqdm(range(len(file_list)), desc=msg):
vectors = com.file_to_vectors(file_list[idx],
n_mels=n_mels,
n_frames=n_frames,
n_fft=n_fft,
hop_length=hop_length,
power=power)
vectors = vectors[: : n_hop_frames, :]
if idx == 0:
data = np.zeros((len(file_list) * vectors.shape[0], dims), float)
data[vectors.shape[0] * idx : vectors.shape[0] * (idx + 1), :] = vectors
return data
result=param["result_directory"],
machine_type=machine_type,
section_name=section_name,
dir_name=dir_name)
decision_result_list = []
print(
"\n============== BEGIN TEST FOR A SECTION ==============")
y_pred = [0. for k in files]
for file_idx, file_path in tqdm(enumerate(files),
total=len(files)):
try:
data = com.file_to_vectors(
file_path,
n_mels=param["feature"]["n_mels"],
n_frames=param["feature"]["n_frames"],
n_fft=param["feature"]["n_fft"],
hop_length=param["feature"]["hop_length"],
power=param["feature"]["power"])
except:
com.logger.error("File broken!!: {}".format(file_path))
y_pred[file_idx] = np.mean(
np.square(data - model.predict(data)))
# store anomaly scores
anomaly_score_list.append(
[os.path.basename(file_path), y_pred[file_idx]])
# store decision results
if y_pred[file_idx] > decision_threshold: