logging.root.removeHandler(handler)
logging.basicConfig(filename="results/training.log",
filemode='w',
level=logging.INFO)
class LogHistory(Callback):
def on_epoch_end(self, batch, logs={}):
logging.info(logs)
model = LSTMAutoencoderV2(batchsize, window_width, encoder_layer_list,
decoder_layer_list)
files = aux_fn.get_all_files(path)
print(['Number of files', len(files)])
data = aux_fn.get_data_from_files(files, sr=sample_rate, ww=window_width)
data = np.expand_dims(data, axis=2)
with tf.Session() as sess:
tf.global_variables_initializer().run()
tf.local_variables_initializer().run()
train_writer = tf.summary.FileWriter(logs_path, sess.graph)
start_time = time.time()
for epoch in range(epochs):
epoch_loss = 0
minibatch = 0
iteration = int((data.shape[0]) / batchsize)
print("Testing folder to proceed:")
print(test_folder)
m = hashlib.md5()
m.update(model_file.encode('utf-8'))
m.update(test_param['test_path'].encode('utf-8'))
response_hash = m.hexdigest()
HIDDEN_LAYER_COUNT = 8
model = aux_fn.gen_model(window_width, LAYER_SIZE, HIDDEN_LAYER_COUNT)
model.load_weights(model_file)
with open(yaml_file, 'r') as infile:
metadata = yaml.load(infile, Loader=yaml.SafeLoader)
files = aux_fn.get_all_files(test_folder)
sig_process = {}
logging.basicConfig(filename='trained_model/inference.log',
filemode='w', level=logging.INFO)
for file in tqdm(files):
file_time = time()
print(file)
logging.info(f' inference on {file}: ')
filename_wav = os.path.basename(file)
for i in range(len(metadata)):
if filename_wav in metadata[i]['mixture_audio_filename']:
file_meta = metadata[i]
break
def main():
log.basicConfig(format="[ %(levelname)s ] %(message)s",
level=log.INFO,
stream=sys.stdout)
args = build_argparser().parse_args()
model_xml = args.model
model_bin = os.path.splitext(model_xml)[0] + ".bin"
with open(args.testparam, 'r') as stream:
test_param = yaml.load(stream, Loader=yaml.FullLoader)
sample_rate = test_param['sample_rate']
rolling_step = test_param['rolling_step'] # this is rolling step
test_folder = test_param['test_path'] + "/audio"
yaml_file = test_param['test_path'] + \
"/meta/mixture_recipes_devtest_gunshot.yaml"
print("Testing folder to proceed:")
print(test_folder)
print(model_bin)
log.info("Creating Inference Engine...")
ie = IECore()
if args.cpu_extension and 'CPU' in args.device:
ie.add_extension(args.cpu_extension, "CPU")
# Read IR
log.info("Loading network files:\n\t{}\n\t{}".format(model_xml, model_bin))
net = IENetwork(model=model_xml, weights=model_bin)
if "CPU" in args.device:
supported_layers = ie.query_network(net, "CPU")
not_supported_layers = [
l for l in net.layers.keys() if l not in supported_layers
]
if len(not_supported_layers) != 0:
log.error(
"Following layers are not supported by the plugin for specified device {}:\n {}"
.format(args.device, ', '.join(not_supported_layers)))
log.error(
"Please try to specify cpu extensions library path in sample's command line parameters using -l "
"or --cpu_extension command line argument")
sys.exit(1)
info_input_blob = None
input_shape_dims = None
feed_dict = {}
for blob_name in net.inputs:
print(len(net.inputs[blob_name].shape))
if len(net.inputs[blob_name].shape) == 3:
input_blob = blob_name
input_shape_dims = 3
elif len(net.inputs[blob_name].shape) == 2:
info_input_blob = blob_name
input_shape_dims = 2
else:
raise RuntimeError(
"Unsupported {}D input layer '{}'. Only 2D and 4D input layers are supported"
.format(len(net.inputs[blob_name].shape), blob_name))
print(len(net.outputs))
assert len(net.outputs) == 1, "Demo supports only single output topologies"
out_blob = next(iter(net.outputs))
log.info("Loading IR to the plugin...")
exec_net = ie.load_network(network=net,
num_requests=1,
device_name=args.device)
# Read and pre-process input image
if input_shape_dims == 3:
batch_size, window_width, num_class = net.inputs[input_blob].shape
elif input_shape_dims == 2:
batch_size, window_width = net.inputs[input_blob].shape
else:
print("check input shape dimension: ")
return
print(net.inputs[input_blob].shape)
files = aux_fn.get_all_files(test_folder)
sig_process = {}
file_meta = None
with open(yaml_file, 'r') as infile:
metadata = yaml.load(infile, Loader=yaml.SafeLoader)
for file in tqdm(files):
file_time = time()
print(file)
logging.info(f' inference on {file}: ')
filename_wav = os.path.basename(file)
for i in range(len(metadata)):
if filename_wav in metadata[i]['mixture_audio_filename']:
file_meta = metadata[i]
break
data = load_data(window_width, file, sample_rate, rolling_step)
if input_shape_dims == 3:
data = np.expand_dims(data, axis=2)
iteration = data.shape[0]
pred = []
for iterate in range(iteration):
print(input_blob)
if input_shape_dims == 3:
feed_dict[input_blob] = data[iterate, :, :]
else:
feed_dict[input_blob] = data[iterate, :]
# inf_start = time.time()
exec_net.start_async(request_id=0, inputs=feed_dict)
if exec_net.requests[0].wait(-1) == 0:
# inf_end = time.time()
# det_time = inf_end - inf_start
res = exec_net.requests[0].outputs[out_blob]
print("result shape", res.shape)
pred.append(list(res))
sig_process[filename_wav] = get_response(data, pred, file_meta,
sample_rate, rolling_step)
tmp_time = time() - file_time
logging.info(
f" \tfile inference time = {tmp_time:.2f} s, {(tmp_time / 117) * 1000:.4f} ms per sample"
)