class KeyboardSensor(object):
def __init__(self, skip_neural=False):
self.sensors = [Button(i) for i in cfg.SENSE_PINS]
self.buffer_size = cfg.window_update_hz * cfg.window_size_s
self.buffer = [0x00] * self.buffer_size
self.idle = True
self.score = 0.5
self.last_NN_outputs = None
if not skip_neural:
self.neuralModel = NeuralNet(model_path=cfg.model_path)
self.neuralModel.load_weights()
def read_sensors(self):
return [s.is_pressed for s in self.sensors]
def get_num_sensors_engaged(self):
""" Returns n sensors in format 0xab, where a is left sum and b is right """
allsensors = self.read_sensors()
left = 4 - sum(allsensors[0:4])
right = 4 - sum(allsensors[4:8])
return hex(left * 16 + right)
def start_sensor_polling(self):
self.score = 0.5
self.kill_logging = threading.Event()
self.logthread = threading.Thread(
target=self.update_buffer) #, daemon=True)
self.logthread.start()
def stop_sensor_polling(self):
self.kill_logging.set()
self.logthread.join()
def update_score_neural(self):
buffer_ints = [int(str(r), 16) for r in self.buffer]
pecking, active = self.neuralModel.Predict(buffer_ints)
# print(pecking, active)
if active > cfg.changes_thresh:
if pecking > cfg.pecking_thresh:
self.score -= cfg.score_inc_dn
else:
self.score += cfg.score_inc_up
self.normalize_score()
self.last_NN_outputs = [pecking, active]
def get_diffs(self, l, r):
last = [l[0], r[0]]
diffs = [0, 0]
for lp, rp in zip(l, r):
if lp != last[0]:
diffs[0] += 1
if rp != last[1]:
diffs[1] += 1
last = [lp, rp]
return diffs
def update_score(self):
ints = [int(str(i), 16) for i in self.buffer]
l = [int(k / 16) for k in ints]
r = [k % 16 for k in ints]
diffs = self.get_diffs(l, r)
# print(l)
# print(r)
# print(sum(l) + sum(r), sum(diffs))
sumcnt = (sum(l) + sum(r)) / 2
if sum(diffs) / 2 < cfg.changes_thresh or sumcnt < cfg.idle_sum_thresh:
self.idle = True
elif sumcnt < cfg.peck_sum_thresh:
self.idle = False
self.score -= cfg.score_inc_dn
else:
self.idle = False
self.score += cfg.score_inc_up
self.normalize_score()
def normalize_score(self):
self.score = 0 if self.score <= 0 else self.score
self.score = 1 if self.score >= 1 else self.score
def get_score(self):
return self.score
def set_score(self, newscore):
self.score = newscore
def update_buffer(self):
buffer_i = 0
time_step_s = 1 / cfg.window_update_hz
next_step_s = time.time() + time_step_s
while not self.kill_logging.is_set():
while time.time() < next_step_s:
pass
self.buffer[buffer_i] = self.get_num_sensors_engaged()
buffer_i += 1
buffer_i %= self.buffer_size
next_step_s += time_step_s
def buffer_dump(self):
return self.buffer
