def get_servo_dataset(filename,
start_index=0,
end_index=None,
conf='/home/ubuntu/settings.json'):
# data = pd.DataFrame.from_csv(filename,encoding='utf8')
data = pd.read_csv(filename,
encoding='utf8',
engine='python',
error_bad_lines=False)
# Outputs
x = []
# Servo ranges from 40-150
servo = []
for i in data.index[start_index:end_index]:
# Don't want noisy data
# if data['servo'][i] < 40 or data['servo'][i] > 150:
# continue
# Append
x.append(deserialize_image(data['image'][i], config=conf))
servo.append(raw_to_cnn(data['servo'][i]))
return x, servo
def visualize_data(filename, width=72, height=48, depth=3, cnn_model=None, conf='/home/ubuntu/settings.json'):
"""
When cnn_model is specified it'll show what the cnn_model predicts (red)
as opposed to what inputs it actually received (green)
"""
data = pd.read_csv(filename, engine='python', error_bad_lines=False)
for i in data.index:
cur_img = data['image'][i]
cur_throttle = int(data['servo'][i])
cur_motor = int(data['motor'][i])
# [1:-1] is used to remove '[' and ']' from string
cur_img_array = deserialize_image(cur_img, config=conf)
y_input = cur_img_array.copy() # NN input
# And then rescale it so we can more easily preview it
cur_img_array = cv2.resize(cur_img_array, (480, 320), interpolation=cv2.INTER_CUBIC)
# Extra debugging info (e.g. steering etc)
cv2.putText(cur_img_array, "frame: %s" % str(i), (5,35), cv2.FONT_HERSHEY_SIMPLEX, 1, 255)
cv2.line(cur_img_array, (240, 300), (240-(90-cur_throttle), 200), (0, 255, 0), 3)
# Motor values
# RGB
cv2.line(cur_img_array, (50, 160), (50, 160-(90-cur_motor)), raw_motor_to_rgb(cur_motor), 3)
# If we wanna visualize our cnn_model
if cnn_model:
y = cnn_model.predict([y_input])
servo_out = cnn_to_raw(y[0])
cv2.line(cur_img_array, (240, 300), (240-(90-int(servo_out)), 200), (0, 0, 255), 3)
# Can determine the motor our with a simple exponential equation
# x = abs(servo_out-90)
# motor_out = (7.64*e^(-0.096*x)) - 1
# motor_out = 90 - motor_out
x_ = abs(servo_out - 90)
motor_out = (7.64*np.e**(-0.096*x_)) - 1
motor_out = int(80 - motor_out) # Only wanna go forwards
cv2.line(cur_img_array, (100, 160), (100, 160-(90-motor_out)), raw_motor_to_rgb(motor_out), 3)
print(motor_out, cur_motor)
# Show frame
# Convert to BGR cause thats how OpenCV likes it
cv2.imshow('frame', cv2.cvtColor(cur_img_array, cv2.COLOR_RGB2BGR))
if cv2.waitKey(0) & 0xFF == ord('q'):
break
def get_servo_dataset(filename, start_index=0, end_index=None):
data = pd.DataFrame.from_csv(filename)
# Outputs
x = []
# Servo ranges from 40-150
servo = []
for i in data.index[start_index:end_index]:
# Don't want noisy data
if data['servo'][i] < 40 or data['servo'][i] > 150:
continue
# Append
x.append(deserialize_image(data['image'][i]))
servo.append(raw_to_cnn(data['servo'][i]))
return x, servo
def visualize_data(filename,
width=70,
height=40,
depth=3,
cnn_model=Model,
conf='./settings.json'):
"""
When cnn_model is specified it'll show what the cnn_model predicts (red)
as opposed to what inputs it actually received (green)
"""
#data = pd.read_csv(filename, engine='python', error_bad_lines=False)
data = pd.read_csv(filename)
model = Model()
for i in data.index:
cur_img = data['image'][i]
cur_throttle = float(data['servo'][i])
cur_motor = float(data['motor'][i])
print(cur_motor)
# [1:-1] is used to remove '[' and ']' from string
cur_img_array = deserialize_image(cur_img, config=conf)
y_input = cur_img_array.copy() # NN input
# And then rescale it so we can more easily preview it
cur_img_array = cv2.resize(cur_img_array, (480, 320),
interpolation=cv2.INTER_CUBIC)
#model.predict(cur_img_array)
#pred_servo = model.predict(cur_img_array)
# Extra debugging info (e.g. steering etc)
cv2.putText(cur_img_array, "frame: %s" % str(i), (5, 35),
cv2.FONT_HERSHEY_SIMPLEX, 1, 255)
cur_throttle = int(1800 * (cur_throttle - 0.15) + 90) #为什么取0.15作为均值
cv2.line(cur_img_array, (240, 300), (240 - (90 - cur_throttle), 200),
(0, 255, 0), 3)
#cv2.line(cur_img_array, (240, 300), (240-(90-cur_throttle), 200), (0, 0, 255), 3)
# Motor values
# RGB
cur_motor = int(1800 * (cur_motor - 0.15) + 90)
cv2.line(cur_img_array, (50, 160), (50, 160 - (90 - cur_motor)),
raw_motor_to_rgb(cur_motor), 3)
# If we wanna visualize our cnn_model
if cnn_model:
#y = cnn_model.predict([pre])
y = model.predict(y_input) #y是float类型的值
#print(type(y))
#print(y)
#servo_out = cnn_to_raw(y) #这里我想知道 cnn_output与raw_output的关系！！！
#print("servo_out:",servo_out)
print("y_angle:", 240 - (90 - int(1800 * (y - 0.15) + 90)))
print("servo_out:{} <---> {}".format(data['servo'][i], y))
cv2.line(cur_img_array, (240, 300),
(240 - (90 - int(1800 * (y - 0.15) + 90)), 200),
(0, 0, 255), 3)
#cv2.line(cur_img_array, (240, 300), (240-(90-int(servo_out)), 200), (0, 0, 255), 3)
#cv2.line(cur_img_array, (240, 300), (240-(90-int(y)), 200), (0, 0, 255), 3)
# Can determine the motor our with a simple exponential equation
# x = abs(servo_out-90)
# motor_out = (7.64*e^(-0.096*x)) - 1
# motor_out = 90 - motor_out
'''
x_ = abs(servo_out - 90)
motor_out = (7.64*np.e**(-0.096*x_)) - 1
motor_out = int(80 - motor_out) # Only wanna go forwards
cv2.line(cur_img_array, (100, 160), (100, 160-(90-motor_out)), raw_motor_to_rgb(motor_out), 3)
print(motor_out, cur_motor)
'''
# Show frame
# Convert to BGR cause thats how OpenCV likes it
cv2.imshow('frame', cv2.cvtColor(cur_img_array, cv2.COLOR_RGB2BGR))
if cv2.waitKey(0) & 0xFF == ord('q'):
break
def visualize_data(filename,
width=70,
height=40,
depth=3,
cnn_model=None,
conf='../settings.json'):
"""
When cnn_model is specified it'll show what the cnn_model predicts (red)
as opposed to what inputs it actually received (green)
"""
data = pd.read_csv(filename)
model = Model()
for i in range(len(data)):
#cur_img = data.image.values[i]
#cur_throttle = float(data.servo.values[i])
#cur_motor = float(data.motor.values[i])
#print(cur_motor)
cur_img = data['image'][i]
#print("cur_img:",cur_img)
cur_throttle = float(data['servo'][i])
#print("cur_throttle:",cur_throttle)
cur_motor = float(data['motor'][i])
#print("cur_motor:",cur_motor)
# [1:-1] is used to remove '[' and ']' from string
cur_img_array = deserialize_image(cur_img, config=conf)
#print("cur_img_array:",cur_img_array)
y_input = cur_img_array.copy() # NN input
#y_input = Image.fromarray(y_input)
#print("y_input:",y_input)
#print("type:",type(y_input))
#print("y_input:",y_input.shape)
# And then rescale it so we can more easily preview it
cur_img_array = cv2.resize(cur_img_array, (40, 70),
interpolation=cv2.INTER_CUBIC)
#print("cur_img_array:",cur_img_array)
#print("cur_img_array:",cur_img_array.shape)
predict_servo = model.predict(y_input)
print("predict_servo:", predict_servo)
# Extra debugging info (e.g. steering etc)
cv2.putText(cur_img_array, "frame: %s" % str(i), (5, 35),
cv2.FONT_HERSHEY_SIMPLEX, 1, 255)
cur_throttle = int(1800 * (cur_throttle - 0.15) + 90)
predict_servo = int(1800 * (predict_servo - 0.15) + 90)
print("yes here")
#cv2.line(cur_img_array, (240, 300), (240-(90-cur_throttle), 200), (0, 255, 0), 3)
cv2.line(y_input, (240, 300), (240 - (90 - predict_servo), 200),
(0, 0, 255), 3)
# Motor values
# RGB
cur_motor = int(1800 * (cur_motor - 0.15) + 90)
cv2.line(cur_img_array, (50, 160), (50, 160 - (90 - cur_motor)),
raw_motor_to_rgb(cur_motor), 3)
# If we wanna visualize our cnn_model
if cnn_model:
y = model.predict([y_input])
#servo_out = cnn_to_raw(y[0])
#cv2.line(y_input, (240, 300), (240-(90-int(servo_out)), 200), (0, 0, 255), 3)
cv2.line(y_input, (240, 300), (240 - (90 - int(y)), 200),
(0, 0, 255), 3)
# Can determine the motor our with a simple exponential equation
# x = abs(servo_out-90)
# motor_out = (7.64*e^(-0.096*x)) - 1
# motor_out = 90 - motor_out
x_ = abs(servo_out - 90)
motor_out = (7.64 * np.e**(-0.096 * x_)) - 1
motor_out = int(80 - motor_out) # Only wanna go forwards
cv2.line(cur_img_array, (100, 160), (100, 160 - (90 - motor_out)),
raw_motor_to_rgb(motor_out), 3)
print(motor_out, cur_motor)
# Show frame
# Convert to BGR cause thats how OpenCV likes it
cv2.imshow('frame', cv2.cvtColor(cur_img_array, cv2.COLOR_RGB2BGR))
if cv2.waitKey(0) & 0xFF == ord('q'):
break