def main():
frames = 0.0
elapsed = time.time()
last_time = time.time()
frame_loop = 0
leftCount = 0
straightCount = 0
rightCount = 0
y_van = 11*win_h/50
print("----- Loading Weights -----")
config = tf.ConfigProto()
config.gpu_options.allow_growth = True # allow dynamic allocation of memory
with tf.Session(config=config) as sess:
sess.run(tf.global_variables_initializer())
if (data_version == "yuv"):
saver = tf.train.import_meta_graph("./PNN_Model{}.meta".format(PNN_VERSION))
else:
saver = tf.train.import_meta_graph("./CNN_Model{}.meta" .format(CNN_VERSION))
saver.restore(sess, tf.train.latest_checkpoint("./"))
for i in range(0,3):
print("On the count of 3: {}" .format(i))
time.sleep(.5)
paused = False
while(True):
if (not paused):
# 800x600 windowed mode
# bbox(x, y, width, height)
vertices = np.array([[0, win_h], [0, 7*win_h/16], [win_w/4, y_van], [3*win_w/4, y_van], [win_w, 7*win_h/16], [win_w, win_h], [5*win_w/8, 11*win_h/16], [3*win_w/8, 11*win_h/16]], np.int32)
image = np.array(ImageGrab.grab(bbox=(win_x, win_y, rect[2], rect[3]))) # grabbing screen into a numpy array // for GTAV
if (data_version == "yuv"):
processed_img = cv2.cvtColor(image, cv2.COLOR_BGR2YUV)
else:
processed_img = cv2.cvtColor(image, cv2.COLOR_BGR2GRAY)
screen = cv2.resize(processed_img, (WIDTH,HEIGHT))
print("loop took {} seconds " .format(time.time()-last_time))
last_time = time.time()
prediction = model.eval({x: screen.reshape(-1, HEIGHT, WIDTH)})[0]
move = list(np.around(prediction))
print("Move: {}" .format(move))
# starting the gas
#straight()
if move == [1,0,0]:
# left()
leftCount += 1
print("Steering: Left")
elif move == [0,1,0]:
# straight()
straightCount += 1
print("Steering: Straight")
elif move == [0,0,1]:
# right()
rightCount += 1
print("Steering: Right")
keys = KeyCheck()
if 'T' in keys:
if paused:
paused = False
time.sleep(.5)
else:
paused = True
ReleaseKey(A)
ReleaseKey(W)
ReleaseKey(D)
time.sleep(.5)
if "Q" in keys:
break
# for counting the number of frames
frames = frames + 1
print("Number of frames: {} " .format(frames))
if cv2.waitKey(25) & 0xFF == ord('q'):
cv2.destroyAllWindows()
final_time = time.time()
print("Average FPS: {} " .format(frames/(final_time - elapsed)))
break
print("Left: {}".format(leftCount))
print("Straight: {}".format(straightCount))
print("Right: {}".format(rightCount))
def right(): ReleaseKey(A) # ReleaseKey(W) PressKey(D)
def straight(): ReleaseKey(A) ReleaseKey(D) PressKey(W)
def left(): ReleaseKey(D) # ReleaseKey(W) PressKey(A)
def main():
frames = 0.0
elapsed = time.time()
last_time = time.time()
frame_loop = 0
y_van = 11 * win_h / 50
for i in range(0, 3):
print("On the count of 3: {}".format(i))
time.sleep(.5)
paused = False
while (True):
if (not paused):
# 800x600 windowed mode
# bbox(x, y, width, height)
vertices = np.array(
[[0, win_h], [0, 7 * win_h / 16], [win_w / 4, y_van],
[3 * win_w / 4, y_van], [win_w, 7 * win_h / 16],
[win_w, win_h], [5 * win_w / 8, 11 * win_h / 16],
[3 * win_w / 8, 11 * win_h / 16]], np.int32)
image = np.array(
ImageGrab.grab(
bbox=(win_x, win_y, rect[2], rect[3]
))) # grabbing screen into a numpy array // for GTAV
grey = cv2.cvtColor(image, cv2.COLOR_BGR2GRAY)
screen = cv2.resize(grey, (WIDTH, HEIGHT))
print("loop took {} seconds ".format(time.time() - last_time))
last_time = time.time()
prediction = model.predict(screen.reshape(1, WIDTH, HEIGHT, 1))[0]
moves = list(np.around(prediction))
print(moves, prediction)
if moves == [1, 0, 0]:
left()
elif moves == [0, 1, 0]:
straight()
elif moves == [0, 0, 1]:
right()
keys = KeyCheck()
if 'T' in keys:
if paused:
paused = False
time.sleep(.5)
else:
pause = True
ReleaseKey(A)
ReleaseKey(W)
ReleaseKey(D)
time.sleep(.5)
# for counting the number of frames
frames = frames + 1
if (frames == 1000):
break
print("Number of frames: {} ".format(frames))
if cv2.waitKey(25) & 0xFF == ord('q'):
cv2.destroyAllWindows()
final_time = time.time()
print("Average FPS: {} ".format(frames / (final_time - elapsed)))
break