def __init__(self, parent):
self.states = {
"model_ready": False,
"model_source": None,
"layers_changed": False,
"training_data_ready": False,
"training_data_files": [],
"test_data_ready": False,
"test_data_files": [],
"settings_locked": False,
}
self.gui_handle = parent
self.status_widgets = []
self.model_data = {"loaded": False}
self.keras_handle = kp.MachineLearning()
def trigger_support_model(self, data, feature_map, trigger_options):
threshold = trigger_options["auto_threshold"]
if not self.trigger_variables.get("initialized", False):
if not hasattr(self, 'keras_process'):
self.kp = keras_p.MachineLearning()
else:
self.kp.clear_model()
trigger_model, t_message = self.kp.load_model("model_trigger.npy")
if not trigger_model["loaded"]:
print(t_message)
return True, None, t_message
else:
print("model loaded")
config = trigger_model["sensor_config"]
feature_list = trigger_model["feature_list"]
frame_settings = trigger_model["frame_settings"]
if frame_settings["collection_mode"] == "support_model":
print("Not supported, changing to continuous")
frame_settings["collection_mode"] = "continuous"
frame_settings["rolling"] = True
self.trigger_process = feat_p.FeatureProcessing(config)
self.trigger_process.set_feature_list(feature_list)
self.trigger_process.set_frame_settings(frame_settings)
self.trigger_variables["initialized"] = True
ml_frame_data = self.trigger_process.feature_extraction(data)
feature_map = ml_frame_data["current_frame"]["feature_map"]
complete = ml_frame_data["current_frame"]["frame_complete"]
score = 0
detected = False
if complete and feature_map is not None:
self.kp.clear_session()
trigger = self.kp.predict(feature_map)[0]
self.kp.clear_session()
score = trigger["label_predictions"]["centered"][0]
if score > threshold:
detected = True
return detected, score, ""
def main():
parser = example_utils.ExampleArgumentParser(num_sens=2)
add_args(parser)
args = parser.parse_args()
if args.model_file_name:
filename = args.model_file_name
else:
print("Not implemented!")
sys.exit(1)
keras_proc = kp.MachineLearning()
model_data = keras_proc.load_model(filename)
print(model_data["message"])
if not model_data["loaded"]:
return False
config = model_data["sensor_config"]
feature_list = model_data["feature_list"]
frame_settings = model_data["frame_settings"]
feature_process = feature_proc.FeatureProcessing(config)
feature_process.set_feature_list(feature_list)
feature_process.set_frame_settings(frame_settings)
example_utils.config_logging(args)
if args.socket_addr:
client = SocketClient(args.socket_addr)
elif args.spi:
client = SPIClient()
else:
port = args.serial_port or example_utils.autodetect_serial_port()
client = UARTClient(port)
info = client.setup_session(config)
interrupt_handler = example_utils.ExampleInterruptHandler()
print("Press Ctrl-C to end session")
client.start_streaming()
init = True
while not interrupt_handler.got_signal:
info, sweep = client.get_next()
if init:
init = False
num_sensors, data_len = sweep.shape
start_x = config.range_interval[0]
stop_x = config.range_interval[1]
x_mm = np.linspace(start_x, stop_x, data_len) * 1000
data = {
"x_mm": x_mm,
"iq_data": sweep,
"env_ampl": abs(sweep),
"sensor_config": config,
"num_sensors": num_sensors,
}
ml_frame_data = feature_process.feature_extraction(data)
feature_map = ml_frame_data["current_frame"]["feature_map"]
complete = ml_frame_data["current_frame"]["frame_complete"]
if complete and feature_map is not None:
predict = keras_proc.predict(feature_map)[0]
prediction_label = predict["prediction"]
print(prediction_label)
print("Disconnecting...")
client.disconnect()
def main():
parser = argparse.ArgumentParser()
parser.add_argument("-s",
"--load-train-set",
dest="server",
help="IP of streaming server",
default="127.0.0.1")
args = parser.parse_args()
filename = "model_data_speaker_sparse.npy"
keras_proc = kp.MachineLearning()
model_data, message = keras_proc.load_model(filename)
print(message, "\n")
if not model_data["loaded"]:
print("Failed to load model!")
sys.exit(1)
try:
client = SocketClient(args.server)
except Exception:
print("Failed to connect to server at {}!\n".format(args.server))
traceback.print_exc()
sys.exit(1)
conf_speaker = model_data["sensor_config"]
conf_speaker.sensor = SENSOR_SPEAKER
feature_list = model_data["feature_list"]
feature_list[0]["sensors"] = [SENSOR_SPEAKER[0]]
feature_list[1]["sensors"] = [SENSOR_SPEAKER[1]]
frame_settings = model_data["frame_settings"]
frame_settings["auto_threshold"] = 1.5
frame_settings["dead_time"] = 30
frame_settings["auto_offset"] = 15
frame_settings["collection_mode"] = "auto_feature_based"
feature_process = feature_proc.FeatureProcessing(conf_speaker)
feature_process.set_feature_list(feature_list)
feature_process.set_frame_settings(frame_settings)
handles = init_demo()
handles["feature_process"] = feature_process
handles["keras_proc"] = keras_proc
# get session config for speaker mode
info_speaker = client.setup_session(conf_speaker)
handles["dist_processors"], handles[
"dist_tags"] = setup_distance_detectors(conf_speaker, info_speaker,
SENSOR_SPEAKER)
try:
client.start_session()
client.stop_session()
except Exception:
print("Failed to start session!")
traceback.print_exc()
sys.exit(1)
demo_mode = "speaker"
if USE_PRESENCE:
# get session config for presence mode
demo_mode = "presence"
conf_presence = presence_detection.get_sensor_config()
conf_presence.sensor = SENSOR_PRESENCE
conf_presence.range_interval = PRESENCE_RANGE
info_presence = client.setup_session(conf_presence)
handles["presence_processor"] = setup_presence_detector(
conf_presence, info_presence)
if USE_PRESENCE:
start_mode = "presence"
else:
start_mode = "speaker"
print("Starting demo in {}-mode!".format(start_mode))
interrupt_handler = utils.ExampleInterruptHandler()
print("Press Ctrl-C to end demo")
client.start_session()
while not interrupt_handler.got_signal:
try:
info, sweep = client.get_next()
if demo_mode == "presence":
new_mode = do_presence_mode(info, sweep, handles)
else:
data = {
"sweep_data": sweep,
"sensor_config": conf_speaker,
"session_info": info_speaker,
}
new_mode = do_speaker_mode(info, data, handles)
# switch between presence and speaker mode
if new_mode != demo_mode:
demo_mode = new_mode
handles["led_handle"].double_flash("#000000")
time.sleep(1)
client.stop_session()
if demo_mode == "presence":
print("Switching to presence mode!\n")
handles["led_handle"].double_flash("#000000")
time.sleep(1)
handles["idle_counts"] = 0
info_presence = client.setup_session(conf_presence)
handles["presence_processor"] = setup_presence_detector(
conf_presence, info_presence)
if handles["play_mode"] == "play":
color = "#00ff00"
else:
color = "#ff0000"
handles["led_handle"].set_color(color,
pos=[3, 4],
brightness=0.1)
else:
print("Switching to speaker mode!\n")
info_speaker = client.setup_session(conf_speaker)
handles["led_handle"].set_color("#00ff00",
pos=[3, 4],
brightness=0.1)
client.start_session()
except Exception:
traceback.print_exc()
break
print("Disconnecting...")
if handles["play_mode"] == "play":
handles["lms_handle"].queue.put("PAUSE")
handles["lms_handle"].stop()
handles["led_handle"].off()
handles["led_handle"].exit()
client.disconnect()
