def main():
for i in list(range(4))[::-1]:
print(i+1)
time.sleep(1)
paused = False
while(True):
if not paused:
# 800x600 windowed mode
screen = grab_screen(region=(0,40,800,640))
last_time = time.time()
screen = cv2.cvtColor(screen, cv2.COLOR_BGR2GRAY)
screen = cv2.resize(screen, (160,120))
# resize to something a bit more acceptable for a CNN
keys = key_check()
output = keys_to_output(keys)
training_data.append([screen,output])
if len(training_data) % 1000 == 0:
print(len(training_data))
np.save(file_name,training_data)
keys = key_check()
if 'T' in keys:
if paused:
paused = False
print('unpaused!')
time.sleep(1)
else:
print('Pausing!')
paused = True
time.sleep(1)
def main(file_name, starting_value):
file_name = file_name
starting_value = starting_value
training_data = []
for i in list(range(4))[::-1]:
print(i+1)
time.sleep(1)
last_time = time.time()
paused = False
print('STARTING!!!')
while(True):
if not paused:
# windowed mode, this is 1920x1080, but you can change this to suit whatever res you're running.
screen = grab_screen(region=(0,40,1920,1120))
last_time = time.time()
# resize to something a bit more acceptable for a CNN
screen = cv2.resize(screen, (480,270))
# run a color convert:
screen = cv2.cvtColor(screen, cv2.COLOR_BGR2RGB)
keys = key_check()
output = keys_to_output(keys)
training_data.append([screen,output])
#print('loop took {} seconds'.format(time.time()-last_time))
last_time = time.time()
## cv2.imshow('window',cv2.resize(screen,(640,360)))
## if cv2.waitKey(25) & 0xFF == ord('q'):
## cv2.destroyAllWindows()
## break
if len(training_data) % 100 == 0:
print(len(training_data))
if len(training_data) == 500:
np.save(file_name,training_data)
print('SAVED')
training_data = []
starting_value += 1
file_name = 'training_data-{}.npy'.format(starting_value)
keys = key_check()
if 'T' in keys:
if paused:
paused = False
print('unpaused!')
time.sleep(1)
else:
print('Pausing!')
paused = True
time.sleep(1)
def main():
last_time = time.time()
for i in list(range(4))[::-1]:
print(i+1)
time.sleep(1)
paused = False
while(True):
if not paused:
# 800x600 windowed mode
screen = grab_screen(region=(0,40,800,640))
print('loop took {} seconds'.format(time.time()-last_time))
last_time = time.time()
screen = cv2.cvtColor(screen, cv2.COLOR_BGR2GRAY)
screen = cv2.resize(screen, (160,120))
prediction = model.predict([screen.reshape(160,120,1)])[0]
print(prediction)
if np.argmax(prediction) == np.argmax(w):
straight()
elif np.argmax(prediction) == np.argmax(s):
reverse()
if np.argmax(prediction) == np.argmax(a):
left()
if np.argmax(prediction) == np.argmax(d):
right()
if np.argmax(prediction) == np.argmax(wa):
forward_left()
if np.argmax(prediction) == np.argmax(wd):
forward_right()
if np.argmax(prediction) == np.argmax(sa):
reverse_left()
if np.argmax(prediction) == np.argmax(sd):
reverse_right()
if np.argmax(prediction) == np.argmax(nk):
no_keys()
keys = key_check()
# p pauses game and can get annoying.
if 'T' in keys:
if paused:
paused = False
time.sleep(1)
else:
paused = True
ReleaseKey(A)
ReleaseKey(W)
ReleaseKey(D)
time.sleep(1)
def main():
for i in list(range(4))[::-1]:
print(i+1)
time.sleep(1)
last_time = time.time()
paused = False
print('STARTING!!!')
while(True):
if not paused:
# 800x600 windowed mode
screen = grab_screen(region=(0,40,1920,1120))
last_time = time.time()
# resize to something a bit more acceptable for a CNN
screen = cv2.resize(screen, (160,90))
# run a color convert:
screen = cv2.cvtColor(screen, cv2.COLOR_BGR2RGB)
keys = key_check()
output = keys_to_output(keys)
training_data.append([screen,output])
#print('loop took {} seconds'.format(time.time()-last_time))
last_time = time.time()
## cv2.imshow('window',cv2.resize(screen,(640,360)))
## if cv2.waitKey(25) & 0xFF == ord('q'):
## cv2.destroyAllWindows()
## break
if len(training_data) % 1000 == 0:
print(len(training_data))
if len(training_data) == 4000:
np.save(file_name,training_data)
for i in range(25):
print('DONE!!!!!!!!!!!!!!!!!!!!!!!!!!!!!')
break
keys = key_check()
if 'T' in keys:
if paused:
paused = False
print('unpaused!')
time.sleep(1)
else:
print('Pausing!')
paused = True
time.sleep(1)
import time
import cv2
from grabscreen import grab_screen
# import numpy as np
# final_image = np.zeros((80, 60, 3), dtype=np.int)
time.sleep(3)
for i in range(10):
# screen = grab_screen(region=(590, 485, 635, 480 + 30))
screen = grab_screen(region=(0, 0, 640, 480 + 40))
screen = cv2.cvtColor(screen, cv2.COLOR_BGR2RGB)
speed_image = screen[485:510, 590:635]
screen_cropped = screen[75:465, 10:630]
screen_cropped_resized = cv2.resize(screen_cropped, (62, 39), 0, 0, 0,
cv2.INTER_AREA)
speed_image_resized = cv2.resize(speed_image, (18, 10), 0, 0, 0,
cv2.INTER_AREA)
combined = screen_cropped_resized
combined[-len(speed_image_resized):,
int(
len(screen_cropped_resized[1]) / 2 -
len(speed_image_resized[1]) / 2):int(
len(screen_cropped_resized[1]) / 2 +
len(speed_image_resized[1]) / 2)] = speed_image_resized
cv2.imwrite("whole_screen-{}.png".format(i), screen)
cv2.imwrite("speed-{}.png".format(i), speed_image)
cv2.imwrite("cropped-{}.png".format(i), screen_cropped)
#TLD Models Parameters
TLD_MODEL_SIZE = (120, 60, 3)
MODEL_NAME = "TLD.model"
TRAFFIC_LIGHT = 9
lightClasses = ["Red", "Yellow", "Green"]
#Model parameters
tldModel = TLDModel()
tldModel.load_weights(MODEL_NAME)
lightClass = None
index = 2000
while(True):
#Capture screen and convert to RGB
screen = grab_screen(SCREEN_REGION)
screen = cv2.cvtColor(screen, cv2.COLOR_BGR2RGB)
#Detect objects using YOLO
boxes, classes, scores = processYolo(screen)
#Classify traffic light colour (if there are any)
if(classes is not None and TRAFFIC_LIGHT in classes):
tlImage, tlPosition = extractLaneTrafficLight(screen, boxes, classes, scores)
prediction = tldModel.predict(np.expand_dims(tlImage, axis=0))[0]
lightClass = np.argmax(prediction)
highlightSeenTrafficLight(screen, tlPosition)
else:
lightClass = None
#Uncomment to show ALL boxes on screen
def main():
last_time = time.time()
vj.open()
time.sleep(1)
for i in list(range(4))[::-1]:
print(i + 1)
time.sleep(1)
model_name = 'steer_augmentation.h5'
model = load_model(model_name)
# Load Yolo
net = cv2.dnn.readNet("yolov3-tiny.weights", "yolov3-tiny.cfg")
classes = []
with open("coco.names.txt", "r") as f:
classes = [line.strip() for line in f.readlines()]
layer_names = net.getLayerNames()
confidence_threshold = 0.6
paused = False
while (True):
if not paused:
'''-------------Screenshot for YOLO and CNN------------------------'''
screen = grab_screen(
region=(0, 40, 800, 640)) # take screen shot of the screen
img = screen #make a copy for YOLOv3
window_width = 800
img = img[:, round(window_width / 4):round(window_width * 3 / 4)]
img = cv2.cvtColor(img, cv2.COLOR_BGR2RGB)
height, width, channels = img.shape
#window_width = width
print("FPS: ", 1 / (time.time() - last_time)) #print FPS
last_time = time.time()
'''-------------------resize and reshape the input image for CNN----------------'''
screen = cv2.cvtColor(screen, cv2.COLOR_BGR2GRAY)
screen = cv2.resize(screen, (160, 120))
prediction = model.predict([screen.reshape(-1, 160, 120, 1)])[0]
print(prediction)
steering_angle = prediction[0]
#if steering_angle > 0.20 or steering_angle <-0.20:
#steering_angle = steering_angle*1.5
throttle = prediction[1]
brake = 0
if throttle >= 0:
throttle = throttle / 2
else:
throttle = -0.25
#-----------------YOLO IMPLEMENTATION---------------------------------
blob = cv2.dnn.blobFromImage(img,
0.00392, (416, 416), (0, 0, 0),
True,
crop=False)
net.setInput(blob)
outs = net.forward(output_layers)
#
class_ids = []
confidences = []
boxes = []
for out in outs:
for detection in out:
scores = detection[5:]
class_id = np.argmax(scores)
confidence = scores[class_id]
if confidence > confidence_threshold:
# Object detected
center_x = int(detection[0] * width)
center_y = int(detection[1] * height)
w = int(detection[2] * width)
h = int(detection[3] * height)
# Rectangle coordinates
x = int(center_x - w / 2)
y = int(center_y - h / 2)
boxes.append([x, y, w, h])
confidences.append(float(confidence))
class_ids.append(class_id)
indexes = cv2.dnn.NMSBoxes(boxes, confidences, 0.5, 0.4)
for i in range(len(boxes)):
if i in indexes:
if (classes[class_ids[i]] == 'car'
or classes[class_ids[i]] == 'truck'
or classes[class_ids[i]] == 'person'
or classes[class_ids[i]] == 'motorbike'
or classes[class_ids[i]] == 'bus'
or classes[class_ids[i]] == 'train'
or classes[class_ids[i]] == 'stop sign'
): # person bicycle car motorbike bus train stop sign
#PressKey(S)
x, y, w, h = boxes[i]
diag_len = np.sqrt((w * w) + (h * h))
#print (diag_len)
if diag_len >= 150: # Detected object is too close so STOP
print("STOP")
throttle = -1
steering_angle = 0
brake = 1
elif diag_len >= 50 and diag_len < 150: # Detected object is near by so SLOW DOWN
print("SLOW")
if throttle >= 0:
throttle = throttle / 2
else:
throttle = throttle - ((1 + throttle) / 2)
else: #Detected object is far
print("JUST DRIVE")
if throttle >= 0:
throttle = throttle / 2
else:
throttle = throttle - ((1 + throttle) / 2)
setJoy_Steer_Throttle_Brake(steering_angle, throttle, brake)
time.sleep(0.0001)
keys = key_check()
# p pauses game and can get annoying.
if 'T' in keys:
if paused:
paused = False
time.sleep(1)
else:
paused = True
steering_angle = 0
throttle = -1
brake = 0
setJoy_Steer_Throttle_Brake(steering_angle, throttle, brake)
time.sleep(1)
stats = []
L_num = 0
while True:
st_time = time.time()
loop_time_1 = time.time()
# DONE: smarter logic about bomb location/placement to kill enemy and bomb timing
# bomb route clipping? (to avoid trying to go to somewhere that will be unavailable because of bomb(s))
# if((L_num%4)!=1):
# getting the window mode screen
screen = grab_screen(region=(anchorHWidthTopLeft, anchorHeightTopLeft,
anchorWidthBotRight, anchorHeightBotRight))
# pixels characters 2
# screen = cv2.resize(screen, (int(WIDTH / 2), int(HEIGTH / 2)))
# screen = cv2.resize(screen, (int(WIDTH / 8), int(HEIGTH / 8)))
loop_time_11 = time.time()
# run a color convert:
screen = cv2.cvtColor(screen, cv2.COLOR_BGR2RGB)
# ret, screen = camera.read()
loop_time_111 = time.time()
getPlayerPosition = GetPlayerPosition(screen, 1)
def thatsover():
for i in list(range(2))[::-1]:
print(i + 1)
time.sleep(1)
paused = False
while (True):
last_time = time.time()
if not paused:
# 640*480 windowed mode
screen = grabscreen.grab_screen(region=(0, 40, 640, 480))
screen, lines = collect_data.process_img(screen)
screen = cv2.resize(screen, (320, 240))
# screen = np.array(screen).reshape(320,240) # screen.reshape(-1,320,240,1)
cv2.imshow('window', screen)
if cv2.waitKey(25) & 0xFF == ord('q'):
cv2.destroyAllWindows()
break
prediction = model.predict([screen.reshape(320, 240, 1)])[0]
b = np.array([])
print(a)
for i in range(9):
if prediction[i] <= 0.4:
b = np.append(b, [0])
print("1")
elif prediction[i] >= 0.4:
b = np.append(b, [1])
print("0")
print(b)
b = b.tolist()
if b == collect_data.w:
directkeys.w()
print("w")
elif b == collect_data.a:
directkeys.a()
print("a")
elif b == collect_data.s:
directkeys.s()
print("s")
elif b == collect_data.d:
directkeys.d()
print("d")
elif b == collect_data.nk:
directkeys.nokey()
print("nokey")
elif b == collect_data.sa:
directkeys.sa()
print("sa")
elif b == collect_data.sd:
directkeys.sd()
print("sd")
elif b == collect_data.wa:
directkeys.wa()
print("wa")
elif b == collect_data.wd:
directkeys.wd()
print("wd")
else:
print("something wrong")
keys = getkeys.key_check()
if 'T' in keys:
if paused:
paused = False
print('unpaused!')
time.sleep(1)
else:
print('Pausing!')
paused = True
time.sleep(1)
print('Loop took {} seconds'.format(time.time() - last_time))
def main():
# last_time = time.time()
for i in list(range(4))[::-1]:
print(i+1)
time.sleep(1)
paused = False
NUM_CLASSES = 90
detection_graph = tf.Graph()
with detection_graph.as_default():
od_graph_def = tf.GraphDef()
with tf.gfile.GFile(PATH_TO_CKPT, 'rb') as fid:
serialized_graph = fid.read()
od_graph_def.ParseFromString(serialized_graph)
tf.import_graph_def(od_graph_def, name='')
label_map = label_map_util.load_labelmap(PATH_TO_LABELS)
categories = label_map_util.convert_label_map_to_categories(label_map, max_num_classes=NUM_CLASSES, use_display_name=True)
category_index = label_map_util.create_category_index(categories)
def load_image_into_numpy_array(image):
(im_width, im_height) = image.size
return np.array(image.getdata()).reshape(
(im_height, im_width, 3)).astype(np.uint8)
IMAGE_SIZE = (12, 8)
with detection_graph.as_default():
with tf.Session(graph=detection_graph) as sess:
while(True):
if not paused:
# 800x600 windowed mode
#screen = np.array(ImageGrab.grab(bbox=(0,40,800,640)))
screen = grab_screen(region=(0,40,800,640))
screen1 = cv2.resize(screen, (800,640))
image_np = cv2.cvtColor(screen1, cv2.COLOR_BGR2RGB)
image_np_expanded = np.expand_dims(image_np, axis=0)
image_tensor = detection_graph.get_tensor_by_name('image_tensor:0')
boxes = detection_graph.get_tensor_by_name('detection_boxes:0')
scores = detection_graph.get_tensor_by_name('detection_scores:0')
classes = detection_graph.get_tensor_by_name('detection_classes:0')
num_detections = detection_graph.get_tensor_by_name('num_detections:0')
(boxes, scores, classes, num_detections) = sess.run(
[boxes, scores, classes, num_detections],
feed_dict={image_tensor: image_np_expanded})
vis_util.visualize_boxes_and_labels_on_image_array(
image_np,
np.squeeze(boxes),
np.squeeze(classes).astype(np.int32),
np.squeeze(scores),
category_index,
use_normalized_coordinates=True,
line_thickness=8)
cv2.imshow('window',image_np)
# print('loop took {} seconds'.format(time.time()-last_time))
# last_time = time.time()
screen = cv2.cvtColor(screen, cv2.COLOR_BGR2GRAY)
screen = cv2.resize(screen, (160,90))
prediction = model.predict([screen.reshape(160,90,1)])[0]
print(prediction)
turn_thresh = .75
fwd_thresh = 0.6
if prediction[1] > fwd_thresh:
straight()
# print("!")
elif prediction[0] > turn_thresh:
left()
# print("<")
elif prediction[2] > turn_thresh and prediction[2] < 0.95 :
right()
# print(">")
else:
straight()
# print("!!!")
keys = key_check()
# p pauses game and can get annoying.
if 'T' in keys:
if paused:
paused = False
time.sleep(1)
else:
paused = True
ReleaseKey(A)
ReleaseKey(W)
ReleaseKey(D)
time.sleep(1)
if cv2.waitKey(25) & 0xFF == ord('q'):
cv2.destroyAllWindows()
break
# ## Helper code
def load_image_into_numpy_array(image):
(im_width, im_height) = image.size
return np.array(image.getdata()).reshape(
(im_height, im_width, 3)).astype(np.uint8)
# Size, in inches, of the output images.
IMAGE_SIZE = (12, 8)
with detection_graph.as_default():
with tf.Session(graph=detection_graph) as sess:
while True:
#screen = cv2.resize(grab_screen(region=(0,40,1280,745)), (WIDTH,HEIGHT))
screen = cv2.resize(grab_screen(region=(0, 40, 1280, 745)),
(1000, 650))
image_np = cv2.cvtColor(screen, cv2.COLOR_BGR2RGB)
# Expand dimensions since the model expects images to have shape: [1, None, None, 3]
image_np_expanded = np.expand_dims(image_np, axis=0)
image_tensor = detection_graph.get_tensor_by_name('image_tensor:0')
# Each box represents a part of the image where a particular object was detected.
boxes = detection_graph.get_tensor_by_name('detection_boxes:0')
# Each score represent how level of confidence for each of the objects.
# Score is shown on the result image, together with the class label.
scores = detection_graph.get_tensor_by_name('detection_scores:0')
classes = detection_graph.get_tensor_by_name('detection_classes:0')
num_detections = detection_graph.get_tensor_by_name(
'num_detections:0')
# Actual detection.
(boxes, scores, classes, num_detections) = sess.run(
def toward_realm(win_location):
pyautogui.keyDown('z')
pyautogui.keyUp('z')
pyautogui.keyDown('w')
realm_location = cv2.imread('realm.png', 0)
location_aq = False #got realm location
w, h = realm_location.shape
is_stuck = 0
win_location = (int(win_location[0]+(win_location[2] - win_location[0])*0.75), win_location[1], win_location[2], int(win_location[1]+(win_location[3] - win_location[1])*0.33))
#win_location = (win_location[0]+20, win_location[1]+20, win_location[2]-20, win_location[3]-20) #add temp
counter = 0
screen = np.array(grab_screen(region=win_location), dtype='uint8')
frame = cv2.cvtColor(screen, cv2.COLOR_BGR2GRAY)
center = (int(frame.shape[0]/2), int(frame.shape[1]/2))
while frame.any():
#print(frame.any())
res = cv2.matchTemplate(frame,realm_location, cv2.TM_SQDIFF)
min_val, max_val, min_loc, max_loc = cv2.minMaxLoc(res)
min_val = min_val/10000000
#print(min_val)
if min_val < 0.7 and not location_aq:
#location_aq = realm_location_get(screen, center, 0)
pyautogui.keyUp('w')
print('location detected..')
location_aq = True
elif location_aq:
if counter%15==0:
seed = random.randrange(0,100)
realm_location_get(screen, center, seed, True)
counter+=1
else:
if is_stuck%100 == 0:
print('im stuck')
stuck_direction = int((is_stuck/100))
if stuck_direction > 0:
if (stuck_direction)%2==0:
for i in range(stuck_direction):
pyautogui.keyDown('a')
pyautogui.keyUp('a')
else:
for i in range(stuck_direction):
pyautogui.keyDown('d')
pyautogui.keyUp('d')
is_stuck+=1
top_left = min_loc
bottom_right = (top_left[0] + w, top_left[1] + h)
cv2.rectangle(frame,top_left, bottom_right, 255, 2)
#cv2.imshow('output', frame)
if cv2.waitKey(25) & 0xFF == ord('q'):
cv2.destroyAllWindows()
break
screen = np.array(grab_screen(region=win_location), dtype='uint8')
frame = cv2.cvtColor(screen, cv2.COLOR_BGR2GRAY)
count_down(0)
print(win_location)
next_image = cv2.imread('next.png', 0)
pet_yard = cv2.imread('petyard.png', 0)
#toward_realm(win_location)
#cv2.namedWindow('output')
#cv2.setMouseCallback('output',mouse_cb)
nexus = False
nexusing = False
tp=False
check=False
while True:
screen = np.array(grab_screen(region=win_location), dtype='uint8')
frame = cv2.cvtColor(screen, cv2.COLOR_BGR2RGB)
game_win, map_win, fame_level, hp_level, mana_level = process_game_frame(frame)
if not nexus:
hp = get_hp(hp_level)
mana = get_mana(mana_level)
print('HP:{} mana:{}'.format(hp, mana))
if hp < 30:
if check_special(hp_level):
print("Entering Dongen")
watch_done(win_location)
time.sleep(2)
else:
def initial_population(pred):
if (pred == 1):
model = neural_network_model()
model.load_weights('my_model_weights.h5')
# [OBS, MOVES]
training_data = []
# all scores:
scores = []
# just the scores that met our threshold:
accepted_scores = []
# iterate through however many games we want:
for game in range(initial_games):
print(game)
score = 0
# moves specifically from this environment:
game_memory = []
# previous observation that we saw
prev_observation = []
# for each frame in 200
PressKey(0x20)
ReleaseKey(0x20)
for frame in range(goal_steps):
# choose random action (0 or 1)
observation = grab_screen(region=(0, 0, 1919, 1079))
observation = cv2.cvtColor(observation, cv2.COLOR_BGR2GRAY)
observation = cv2.resize(observation, (256, 144))
done = IsDone()
reward = 1
action = 0
if (pred == 1):
observation = observation.reshape(-1, HEIGHT, WIDTH, 1)
action = np.argmax(model.predict(observation))
#print(model.predict(observation))
print(action)
if action == 1:
PressKey(0x26)
ReleaseKey(0x26)
if action == 2:
PressKey(0x28)
ReleaseKey(0x28)
elif (pred == 2):
if win32api.GetAsyncKeyState(win32con.VK_UP) != 0:
action = 1
if win32api.GetAsyncKeyState(win32con.VK_DOWN) != 0:
action = 2
if win32api.GetAsyncKeyState(0x51) != 0:
done = True
else:
action = random.randrange(0, 2)
if action == 1:
PressKey(0x26)
ReleaseKey(0x26)
if action == 2:
PressKey(0x28)
ReleaseKey(0x28)
if len(prev_observation) > 0:
game_memory.append([prev_observation, action])
prev_observation = observation
score += reward
if done:
print(score)
time.sleep(2)
break
if frame == goal_steps: break
if score >= score_requirement:
accepted_scores.append(score)
for data in game_memory:
training_data.append([data[0], action])
scores.append(score)
if training_data:
try:
prev_data = list(np.load('saved.npy', allow_pickle=True))
prev_data.append(training_data)
except Exception as e:
print(e)
# just in case you wanted to reference later
training_data_save = np.array(training_data)
np.save('saved.npy', training_data_save)
# some stats here, to further illustrate the neural network magic!
print('Average accepted score:', mean(accepted_scores))
print('Median score for accepted scores:', median(accepted_scores))
print(Counter(accepted_scores))
return training_data
# from grabscreen import grab_screen
# screen = cv2.resize(grab_screen(region=(0,40,1280,745)), (800,450))
# image_np = cv2.cvtColor(screen, cv2.COLOR_BGR2RGB)
# show_inference(detection_model, image_np)
#
from grabscreen import grab_screen
import cv2
import time
last_time = time.time()
while True:
screen = cv2.resize(grab_screen(region=(0, 40, 1000, 700)), (800, 450))
image_np = cv2.cvtColor(screen, cv2.COLOR_BGR2RGB)
output_dict = run_inference_for_single_image(detection_model, image_np)
vis_util.visualize_boxes_and_labels_on_image_array(
image_np,
output_dict['detection_boxes'],
output_dict['detection_classes'],
output_dict['detection_scores'],
category_index,
instance_masks=output_dict.get('detection_masks_reframed', None),
use_normalized_coordinates=True,
line_thickness=8)
cv2.imshow('window', image_np)
def main(file_name, starting_value):
file_name = file_name
starting_value = starting_value
training_data = []
for i in list(range(4))[::-1]:
print(i+1)
time.sleep(1)
last_time = time.time()
paused = False
print('STARTING!!!')
intI = 0
while(True):
if(paused == True):
time.sleep(1)
paused = False
# DONE: find the proper anchor
while paused==False:
print("========================")
# print("another frame")
last_time = time.time()
if (keyboard.is_pressed('n') == True):
paused = True
# getting the window mode screen
screen = grab_screen(region=(anchorHWidthTopLeft,anchorHeightTopLeft,
anchorWidthBotRight,anchorHeightBotRight))
last_time = time.time()
# pixels characters 2
# resize to something a bit more acceptable for a CNN
screen = cv2.resize(screen, (int(WIDTH/2), int(HEIGTH/2)))
# run a color convert:
screen = cv2.cvtColor(screen, cv2.COLOR_BGR2RGB)
# which key is pressed
key = what_keys_is_pressed()
print("key:"+str(key))
output = key
if(key==-1):
print("if(key==-1):")
else:
if(1):
intI = intI + 1
print("intI:"+str(intI))
training_data.append([screen,output])
if len(training_data) % 100 == 0:
print("len(training_data):" + str(len(training_data)))
if len(training_data) == 500:
np.save(file_name, training_data)
print('SAVED')
training_data = []
starting_value += 1
file_name = './phase-1/training_data-{}.npy'.format(starting_value)
#print('loop took {} seconds'.format(time.time()-last_time))
last_time = time.time()
# print(type(screen))
# too see what is captured
cv2.imshow('screen',cv2.cvtColor(screen, cv2.COLOR_BGR2RGB))
if cv2.waitKey(25) & 0xFF == ord('t'):
cv2.destroyAllWindows()
break
obstacle_height = 0
should_crouch = False
jumpdist = 175
is_crouching = False
MAX_SPEED_TIME = 118
DINO_WALKING_HEIGHT = 110
while (True):
leftest = 1000
if (time.time() - second >= 1
and time.time() - start_time < MAX_SPEED_TIME):
jumpdist += 1.7
second = time.time()
# jump_dist = 170 + (time.time() - start_time) *1.80
pts = []
scr = grab_screen(region=(75, 250, 750, 450))
# scr = cv2.cvtColor(scr, cv2.COLOR_BGR2RGB)
scr_gray = cv2.cvtColor(scr, cv2.COLOR_BGR2GRAY)
res_dino = cv2.matchTemplate(scr_gray, dino, cv2.TM_CCOEFF_NORMED)
threshold = 0.8
loc_dino = np.where(res_dino >= threshold)
for pt in zip(*loc_dino[::-1]):
cv2.rectangle(scr, pt, (pt[0] + w_dino, pt[1] + h_dino + 9),
(50, 205, 50), 1)
dinoX = pt[0] + w_dino
dinoH = pt[1]
is_crouching = False
for cactus in cacti:
res = cv2.matchTemplate(scr_gray, cactus, cv2.TM_CCOEFF_NORMED)
return masked_img
LD = Lane_Detector()
while (True):
tic = time.time()
# First Try - using PIL ImageGrab method - gives less frame rate
# orig_screengrab = np.array(ImageGrab.grab(bbox=(0,40,640,520)))
# Grabbing + Displaying takes around 0.15 secs per frame of size(480,640)
# Capture Region based on resolution
# 40px to allow for the window title bar, bottom = 480 + 40 = 520
capture_region = (0, 40, 640, 520)
orig_screengrab = grab_screen(capture_region)
# Grabbing + Displaying takes around 0.03 secs per frame of size(480,640)
mask_birdeye, mask_1ch, lane_masked_image, roi_extracted_image, orig_image = LD.process_screengrab(
orig_screengrab, capture_region)
#print(processed_screengrab.shape)
cv2.imshow('Original ScreeGrab',
cv2.cvtColor(orig_screengrab, cv2.COLOR_RGB2BGR))
cv2.imshow('Region of Interest Extracted',
cv2.cvtColor(roi_extracted_image, cv2.COLOR_RGB2BGR))
cv2.imshow('Lane Masked Image',
cv2.cvtColor(lane_masked_image, cv2.COLOR_RGB2BGR))
cv2.imshow('Binary Mask', mask_1ch)
cv2.imshow('Mask after Bird Eye Transform', mask_birdeye)
ReleaseKey(D)
def slow_ya_roll():
ReleaseKey(W)
ReleaseKey(A)
ReleaseKey(D)
for i in list(range(4))[::-1]:
print(i+1)
time.sleep(1)
last_time = time.time()
while True:
screen = grab_screen(region=(0,40,800,640))
print('Frame took {} seconds'.format(time.time()-last_time))
last_time = time.time()
new_screen,original_image, m1, m2 = process_img(screen)
#cv2.imshow('window', new_screen)
cv2.imshow('window2',cv2.cvtColor(original_image, cv2.COLOR_BGR2RGB))
if m1 < 0 and m2 < 0:
right()
elif m1 > 0 and m2 > 0:
left()
else:
straight()
#cv2.imshow('window',cv2.cvtColor(screen, cv2.COLOR_BGR2RGB))
if cv2.waitKey(25) & 0xFF == ord('q'):
def main():
for i in list(range(4))[::-1]:
print(i + 1)
time.sleep(1)
paused = False
mode_choice = 0
test_data = []
test_data_2 = []
test_data_3 = []
while (True):
if not paused:
if len(test_data) == 0:
screen = grab_screen(region=(0, 40, GAME_WIDTH,
GAME_HEIGHT + 40))
screen = cv2.cvtColor(screen, cv2.COLOR_BGR2RGB)
screen = cv2.resize(screen, (WIDTH, HEIGHT))
keys = key_check()
output = keys_to_output(keys)
test_data = [screen, output]
elif len(test_data_2) == 0:
screen = grab_screen(region=(0, 40, GAME_WIDTH,
GAME_HEIGHT + 40))
screen = cv2.cvtColor(screen, cv2.COLOR_BGR2RGB)
screen = cv2.resize(screen, (WIDTH, HEIGHT))
keys = key_check()
output = keys_to_output(keys)
test_data_2 = test_data
test_data = [screen, output]
elif len(test_data_3) == 0:
screen = grab_screen(region=(0, 40, GAME_WIDTH,
GAME_HEIGHT + 40))
screen = cv2.cvtColor(screen, cv2.COLOR_BGR2RGB)
screen = cv2.resize(screen, (WIDTH, HEIGHT))
keys = key_check()
output = keys_to_output(keys)
test_data_3 = test_data_2
test_data_2 = test_data
test_data = [screen, output]
else:
screen = grab_screen(region=(0, 40, GAME_WIDTH,
GAME_HEIGHT + 40))
screen = cv2.cvtColor(screen, cv2.COLOR_BGR2RGB)
screen = cv2.resize(screen, (WIDTH, HEIGHT))
img1 = screen
img2 = test_data[0]
img3 = test_data_2[0]
img4 = test_data_3[0]
last_input = test_data[1]
_2nd_last_input = test_data_2[1]
_3rd_last_input = test_data_3[1]
font = cv2.FONT_HERSHEY_SIMPLEX
last = []
if last_input == [1, 0, 0, 0, 0, 0, 0, 0, 0]:
last = 'W'
elif last_input == [0, 1, 0, 0, 0, 0, 0, 0, 0]:
last = 'SS'
elif last_input == [0, 0, 1, 0, 0, 0, 0, 0, 0]:
last = 'AAA'
elif last_input == [0, 0, 0, 1, 0, 0, 0, 0, 0]:
last = 'DDDD'
elif last_input == [0, 0, 0, 0, 1, 0, 0, 0, 0]:
last = 'WAWAWA'
elif last_input == [0, 0, 0, 0, 0, 1, 0, 0, 0]:
last = 'WDWDWDWD'
elif last_input == [0, 0, 0, 0, 0, 0, 1, 0, 0]:
last = 'SASASASASA'
elif last_input == [0, 0, 0, 0, 0, 0, 0, 1, 0]:
last = 'SDSDSDSDSDSD'
elif last_input == [0, 0, 0, 0, 0, 0, 0, 0, 1]:
last = 'NKNKNKNKNKNKNK'
cv2.putText(img2, str(last), (10, 10), font, 1, (0, 0, 255), 3,
cv2.LINE_AA)
if _2nd_last_input == [1, 0, 0, 0, 0, 0, 0, 0, 0]:
last = 'W'
elif _2nd_last_input == [0, 1, 0, 0, 0, 0, 0, 0, 0]:
last = 'SS'
elif _2nd_last_input == [0, 0, 1, 0, 0, 0, 0, 0, 0]:
last = 'AAA'
elif _2nd_last_input == [0, 0, 0, 1, 0, 0, 0, 0, 0]:
last = 'DDDD'
elif _2nd_last_input == [0, 0, 0, 0, 1, 0, 0, 0, 0]:
last = 'WAWAWA'
elif _2nd_last_input == [0, 0, 0, 0, 0, 1, 0, 0, 0]:
last = 'WDWDWDWD'
elif _2nd_last_input == [0, 0, 0, 0, 0, 0, 1, 0, 0]:
last = 'SASASASASA'
elif _2nd_last_input == [0, 0, 0, 0, 0, 0, 0, 1, 0]:
last = 'SDSDSDSDSDSD'
elif _2nd_last_input == [0, 0, 0, 0, 0, 0, 0, 0, 1]:
last = 'NKNKNKNKNKNKNK'
cv2.putText(img3, str(last), (10, 10), font, 1, (0, 0, 255), 3,
cv2.LINE_AA)
if _3rd_last_input == [1, 0, 0, 0, 0, 0, 0, 0, 0]:
last = 'W'
elif _3rd_last_input == [0, 1, 0, 0, 0, 0, 0, 0, 0]:
last = 'SS'
elif _3rd_last_input == [0, 0, 1, 0, 0, 0, 0, 0, 0]:
last = 'AAA'
elif _3rd_last_input == [0, 0, 0, 1, 0, 0, 0, 0, 0]:
last = 'DDDD'
elif _3rd_last_input == [0, 0, 0, 0, 1, 0, 0, 0, 0]:
last = 'WAWAWA'
elif _3rd_last_input == [0, 0, 0, 0, 0, 1, 0, 0, 0]:
last = 'WDWDWDWD'
elif _3rd_last_input == [0, 0, 0, 0, 0, 0, 1, 0, 0]:
last = 'SASASASASA'
elif _3rd_last_input == [0, 0, 0, 0, 0, 0, 0, 1, 0]:
last = 'SDSDSDSDSDSD'
elif _3rd_last_input == [0, 0, 0, 0, 0, 0, 0, 0, 1]:
last = 'NKNKNKNKNKNKNK'
cv2.putText(img4, str(last), (10, 10), font, 1, (0, 0, 255), 3,
cv2.LINE_AA)
vis0 = np.concatenate((img1, img2), axis=1)
vis1 = np.concatenate((img3, img4), axis=1)
final_vis = np.concatenate((vis0, vis1), axis=0)
final_vis = cv2.resize(final_vis, (480, 270))
prediction = model.predict(
[final_vis.reshape(WIDTH, HEIGHT, 3)])[0]
prediction = np.array(
prediction) # * np.array([1.12, 1, 1, 1, 1, 1, 1, 1, 0.2])
print(prediction)
# div = 4
# wplus = prediction[0] + (prediction[4] / div) + (prediction[5] / div)
# print(wplus)
# splus = prediction[1] + (prediction[6] / div) + (prediction[7] / div)
# print(splus)
# aplus = prediction[2] + (prediction[6] / div) + (prediction[4] / div)
# print(aplus)
# dplus = prediction[3] + (prediction[5] / div) + (prediction[7] / div)
# print(dplus)
# waplus = prediction[4] + (prediction[0] / div) + (prediction[2] / div)
# print(waplus)
# wdplus = prediction[5] + (prediction[0] / div) + (prediction[3] / div)
# print(wdplus)
# saplus = prediction[6] + (prediction[2] / div) + (prediction[1] / div)
# print(saplus)
# sdplus = prediction[7] + (prediction[1] / div) + (prediction[3] / div)
# print(sdplus)
#
# mode_choice_v2 = np.argmax([wplus, splus, aplus, dplus, waplus, wdplus, saplus, sdplus])
# print(mode_choice_v2)
mode_choice = np.argmax(prediction)
print(mode_choice)
if mode_choice == 0:
straight()
choice_picked = 'straight'
print(choice_picked)
elif mode_choice == 1:
reverse()
choice_picked = 'reverse'
print(choice_picked)
elif mode_choice == 2:
left()
choice_picked = 'left'
print(choice_picked)
elif mode_choice == 3:
right()
choice_picked = 'right'
print(choice_picked)
elif mode_choice == 4:
forward_left()
choice_picked = 'forward+left'
print(choice_picked)
elif mode_choice == 5:
forward_right()
choice_picked = 'forward+right'
print(choice_picked)
elif mode_choice == 6:
reverse_left()
choice_picked = 'reverse+left'
print(choice_picked)
elif mode_choice == 7:
reverse_right()
choice_picked = 'reverse+right'
print(choice_picked)
elif mode_choice == 8:
no_keys()
choice_picked = 'nokeys'
print(choice_picked)
keys = key_check()
output = keys_to_output(keys)
test_data_3 = test_data_2
test_data_2 = test_data
test_data = [screen, output]
keys = key_check()
if 'T' in keys:
if paused:
paused = False
time.sleep(1)
print("Unpaused")
else:
paused = True
ReleaseKey(A)
ReleaseKey(W)
ReleaseKey(D)
time.sleep(1)
print("Paused")
def main(file_name, starting_value):
file_name = file_name
starting_value = starting_value
training_data = []
for i in list(range(4))[::-1]:
print(i+1)
time.sleep(1)
stop = False
print('STARTING!!!')
intI = 0
# supposed to be at least not a dead player at the start
numbersOFDeathInLastSeconds = 0
p1scoreAS, p1killAS, p1deathAS = getScoreKillsDeaths()
# AS: At Start
print("at start p1scoreAS, p1killAS, p1deathAS", p1scoreAS, p1killAS, p1deathAS)
while(True):
print("========================")
numberOfKeys = len(all_keys)
print("numberOfKeys",numberOfKeys)
roundEnded = False
i = 0
# screenshotTaken = []
# keyIssued = []
if(stop == True):
break
if(starting_value > 100):
exit()
game_data = []
while(numbersOFDeathInLastSeconds==1):
# 1 or >1 if a players in the lasts seconds
tmpOffset = int(0x455C9C)
print("tmpOffset", '%s' % hex(tmpOffset))
ReadProcessMemory(ph, c.c_void_p(tmpOffset), buff, bufferSize, c.byref(bytesRead))
numbersOFDeathInLastSeconds = unpack('I', buff)[0]
print("numbersOFDeathInLastSeconds", numbersOFDeathInLastSeconds)
while(roundEnded == False):
# choosedKey = random.randint(numberOfKeys)
choosedKey = random.randint(300000) % 6
print("choosedKey",choosedKey)
# getting the window mode screen
screen = grab_screen(region=(anchorHWidthTopLeft, anchorHeightTopLeft,
anchorWidthBotRight, anchorHeightBotRight))
# pixels characters 2
# resize to something a bit more acceptable for a CNN
screen = cv2.resize(screen, (int(WIDTH / 2), int(HEIGTH / 2)))
# run a color convert:
screen = cv2.cvtColor(screen, cv2.COLOR_BGR2RGB)
timePress = ( 0.05 + 0.01 * random.randint(7) ) / SPEEDHACK_SPEED
# e = [1,0,0,0,0,0]
# d = [0,1,0,0,0,0]
# s = [0,0,1,0,0,0]
# f = [0,0,0,1,0,0]
# ctrl = [0,0,0,0,1,0]
# none = [0,0,0,0,0,1]
# all_keys = [e,d,s,f,ctrl,none]
if(choosedKey == 0):
keyboard.press('e')
time.sleep(timePress)
keyboard.release('e')
if(choosedKey == 1):
keyboard.press('d')
time.sleep(timePress)
keyboard.release('d')
if(choosedKey == 2):
keyboard.press('s')
time.sleep(timePress)
keyboard.release('s')
if(choosedKey == 3):
keyboard.press('f')
time.sleep(timePress)
keyboard.release('f')
if(choosedKey == 4):
keyboard.press('ctrl')
time.sleep(timePress)
keyboard.release('ctrl')
if (choosedKey == 5):
print("none")
# debugging to test before firing the script for hours
# time.sleep(timePress)
if (keyboard.is_pressed('p') == True):
stop = True
keyboard.release('ctrl')
keyboard.release('e')
keyboard.release('d')
keyboard.release('s')
keyboard.release('f')
break
# screenshotTaken.append(screen)
# keyIssued.append(choosedKey)
game_data.append([screen, choosedKey])
# TODO: check when player2 is dead
# TODO: check for variable set to one in memory
# when the round is restarting/ending
# 1 if a players in the lasts seconds
tmpOffset = int(0x4440B0)
print("tmpOffset", '%s' % hex(tmpOffset))
ReadProcessMemory(ph, c.c_void_p(tmpOffset), buff, bufferSize, c.byref(bytesRead))
aplayerDown = unpack('I', buff)[0]
print("playerDown", aplayerDown)
# 1 or >1 if a players in the lasts seconds
tmpOffset = int(0x455C9C)
print("tmpOffset", '%s' % hex(tmpOffset))
ReadProcessMemory(ph, c.c_void_p(tmpOffset), buff, bufferSize, c.byref(bytesRead))
numbersOFDeathInLastSeconds = unpack('I', buff)[0]
print("numbersOFDeathInLastSeconds", numbersOFDeathInLastSeconds)
if(numbersOFDeathInLastSeconds!=0):
p1scoreNew, p1killNew, p1deathNew = getScoreKillsDeaths()
print("p1scoreNew, p1killNew, p1deathNew", p1scoreNew, p1killNew, p1deathNew)
# drop the current capture if p1 and p2 died
if(numbersOFDeathInLastSeconds==1):
print(">>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>")
roundEnded = True
# Mean the controlled player kill count changed
if(p1killNew!=p1killAS):
file_name = './phase-1-bruteforce/training_data-{}.npy'.format(starting_value)
np.save(file_name, game_data)
print('SAVED')
starting_value += 1
keyboard.press('tab')
time.sleep(0.01/SPEEDHACK_SPEED)
keyboard.release('tab')
time.sleep(0.01/SPEEDHACK_SPEED)
# updating infos about the game state, a break is required to work properly
p1killAS = p1killNew
p1kdeathAS = p1deathNew
p1scoreAS = p1killNew
pass
break
else:
print("<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<")
# # p1 and p2 are dead (including both killed themself)
# if(numbersOFDeathInLastSeconds>1):
# roundEnded = True
# file_name = './phase-1-bruteforce/training_data-{}.npy'.format(starting_value)
# np.save(file_name, game_data)
# print('SAVED')
# starting_value += 1
#
# keyboard.press('tab')
# time.sleep(0.01/SPEEDHACK_SPEED)
# keyboard.release('tab')
# time.sleep(0.01/SPEEDHACK_SPEED)
#
# # updating infos about the game state, a break is required to work properly
# p1killAS = p1killNew
# p1kdeathAS = p1deathNew
# p1scoreAS = p1killNew
#
# break
# stop the recording if it is too long (and kill the player ?)
if(i == 50000):
roundEnded = True
break
i+=1
screen = get_screen()
obj_locations = get_objects_locations(screen)
#Get center tuples
ball_center = tuple(np.round(obj_locations[:2]).astype(int))
bar1_center = tuple(np.round(obj_locations[2:4]).astype(int))
bar2_center = tuple(np.round(obj_locations[4:6]).astype(int))
#Draw object locations on the image
cv2.circle(screen, ball_center, 5, 128, -1)
cv2.circle(screen, bar1_center, 5, 128, -1)
cv2.circle(screen, bar2_center, 5, 128, -1)
return screen, obj_locations
if __name__ == "__main__":
print("Fit pong screen into the window:")
while True:
screen = grab_screen(region=(650,300,950,600))
cv2.imshow("Capture Screen", screen)
#Pass next frame every 25ms
#Exit when 'q' is pressed
if cv2.waitKey(25) == ord('q'):
cv2.destroyAllWindows()
break
window_size = (320, 104, 704, 448) #384,344 192,172 96,86
blood_window = (60, 91, 280, 562)
for i in list(range(wait_time))[::-1]:
print(i + 1)
time.sleep(1)
last_time = time.time()
while (True):
#printscreen = np.array(ImageGrab.grab(bbox=(window_size)))
#printscreen_numpy = np.array(printscreen_pil.getdata(),dtype='uint8')\
#.reshape((printscreen_pil.size[1],printscreen_pil.size[0],3))
#pil格式耗时太长
screen_gray = cv2.cvtColor(grabscreen.grab_screen(blood_window),
cv2.COLOR_BGR2GRAY) #灰度图像收集
# screen_reshape = cv2.resize(screen_gray,(96,86))
self_blood = self_blood_count(screen_gray)
boss_blood = boss_blood_count(screen_gray)
cv2.imshow('window1', screen_gray)
#cv2.imshow('window3',printscreen)
#cv2.imshow('window2',screen_reshape)
#测试时间用
print('loop took {} seconds'.format(time.time() - last_time))
last_time = time.time()
if cv2.waitKey(5) & 0xFF == ord('q'):
break
ReleaseKey(D)
def slow_ya_roll():
ReleaseKey(W)
ReleaseKey(A)
ReleaseKey(D)
for i in list(range(4))[::-1]:
print(i + 1)
time.sleep(1)
last_time = time.time()
while True:
screen = grab_screen(region=(0, 40, 800, 640))
print('Frame took {} seconds'.format(time.time() - last_time))
last_time = time.time()
new_screen, original_image, m1, m2 = process_img(screen)
#cv2.imshow('window', new_screen)
cv2.imshow('window2', cv2.cvtColor(original_image, cv2.COLOR_BGR2RGB))
if m1 < 0 and m2 < 0:
right()
elif m1 > 0 and m2 > 0:
left()
else:
straight()
#cv2.imshow('window',cv2.cvtColor(screen, cv2.COLOR_BGR2RGB))
if cv2.waitKey(25) & 0xFF == ord('q'):
click_point.append(x/capturing_resize[0]);
click_point.append(y/capturing_resize[1]);
for i in range(len(detected_items)):
if isPointInsideBbox(click_point, detected_items[i]) == True:
print("Inside")
add_tracker(image_np,detected_items[i])
cv2.namedWindow('window')
cv2.setMouseCallback('window',track_selected)
with detection_graph.as_default():
with tf.Session(graph=detection_graph) as sess:
while True:
state_right = win32api.GetKeyState(0x02)
if state_right == 0 or state_right == 1:
screen = cv2.resize(grab_screen(region=(capturing_coordinates)), (capturing_resize))
image_np = cv2.cvtColor(screen, cv2.COLOR_BGR2RGB)
# Expand dimensions since the model expects images to have shape: [1, None, None, 3]
image_np_expanded = np.expand_dims(image_np, axis=0)
image_tensor = detection_graph.get_tensor_by_name('image_tensor:0')
# Each box represents a part of the image where a particular object was detected.
boxes = detection_graph.get_tensor_by_name('detection_boxes:0')
# Each score represent how level of confidence for each of the objects.
# Score is shown on the result image, together with the class label.
scores = detection_graph.get_tensor_by_name('detection_scores:0')
classes = detection_graph.get_tensor_by_name('detection_classes:0')
num_detections = detection_graph.get_tensor_by_name('num_detections:0')
# Actual detection.
(boxes, scores, classes, num_detections) = sess.run(
[boxes, scores, classes, num_detections],
feed_dict={image_tensor: image_np_expanded})
tf.import_graph_def(od_graph_def, name='')
label_map = label_map_util.load_labelmap(PATH_TO_LABELS)
categories = label_map_util.convert_label_map_to_categories(label_map, max_num_classes=NUM_CLASSES, use_display_name=True)
category_index = label_map_util.create_category_index(categories)
def load_image_into_numpy_array(image):
(im_width, im_height) = image.size
return np.array(image.getdata()).reshape(
(im_height, im_width, 3)).astype(np.uint8)
IMAGE_SIZE = (12, 8)
with detection_graph.as_default():
with tf.Session(graph=detection_graph) as sess:
while True:
screen = cv2.resize(grab_screen(region=(0,40,800,640)), (800,640))
image_np = cv2.cvtColor(screen, cv2.COLOR_BGR2RGB)
image_np_expanded = np.expand_dims(image_np, axis=0)
image_tensor = detection_graph.get_tensor_by_name('image_tensor:0')
boxes = detection_graph.get_tensor_by_name('detection_boxes:0')
scores = detection_graph.get_tensor_by_name('detection_scores:0')
classes = detection_graph.get_tensor_by_name('detection_classes:0')
num_detections = detection_graph.get_tensor_by_name('num_detections:0')
(boxes, scores, classes, num_detections) = sess.run(
[boxes, scores, classes, num_detections],
feed_dict={image_tensor: image_np_expanded})
vis_util.visualize_boxes_and_labels_on_image_array(
image_np,
np.squeeze(boxes),
np.squeeze(classes).astype(np.int32),
np.squeeze(scores),
def main():
print("Loaded model: " + MODEL_NAME)
# 4 second countdown
for i in list(range(4))[::-1]:
print(i+1)
time.sleep(1)
last_time = time.time()
time_per_frame = 0
iterations = 0
paused = False
while(True):
if not paused:
screen = grab_screen(region=(0, 0, 640, 480))
screen = cv2.cv2.cvtColor(screen, cv2.COLOR_BGR2GRAY)
screen = cv2.resize(screen, (64, 48))
cv2.namedWindow('input screen')
cv2.moveWindow('input screen', 800,30)
cv2.imshow('input screen', screen)
# taking game frame as input and predicting a one-hot array at that exact frame
img = np.array(screen).reshape(-1,64,48,1) # acnet
# x2 = np.array(screen) # resnet and densenet
# img = np.repeat(x2[..., np.newaxis], 3, -1).reshape(-1, 64, 48, 3) # resnet and densenet
pred = model.predict(np.array(img))
y = list(np.around(pred[0]))
print(y)
if y == [0, 1, 0]:
straight()
elif y == [0, 0, 1]:
right()
elif y == [1, 0, 0]:
left()
# press G to pause the script
keys = key_check()
if 'G' in keys:
if paused:
paused = False
time.sleep(1)
else:
paused = True
ReleaseKey(A)
ReleaseKey(W)
ReleaseKey(D)
time.sleep(1)
iterations+=1
time_per_frame += float(format(time.time() - last_time))
last_time = time.time()
# focus opencv window and press Q to quit
if (cv2.waitKey(25) & 0xFF == ord('q')):
stop()
cv2.destroyAllWindows()
print('Average FPS:', iterations/time_per_frame)
break
print("Starting !!!")
time.sleep(1)
print("1")
time.sleep(1)
print("2")
time.sleep(1)
print("3")
time.sleep(1)
print("4")
print("GO go go go RECORDING STARTED!!!!")
frame = 0
skip = 2
while (True):
current_view = grab_screen([5, 20, 800, 600])
reduced_view = cv2.resize(current_view, (shape[1], shape[0]),
interpolation=cv2.INTER_AREA)
instant_move = key_check()
frame += 1
if (frame % skip == 0 and verify_differnce(instant_move, 2000)):
train_data_collect(reduced_view, instant_move)
frame = 0
cv2.imshow('rv', reduced_view)
if cv2.waitKey(1) == 27:
break
print(moves_count, " ( ", len(X_train[0]), " ) ")
cv2.destroyAllWindows()
cv2.waitKey(1)
save_train_data("second_batch")
#ReleaseKey(A)
def right():
PressKey(D)
ReleaseKey(A)
ReleaseKey(W)
#ReleaseKey(D)
for i in list(range(4))[::-1]:
print(i+1)
time.sleep(1)
while True:
screen = grab_screen(region=(0,40,800,640))
print('Frame took {} seconds'.format(time.time()-last_time))
last_time = time.time()
new_screen,original_image, m1, m2 = process_img(screen)
#cv2.imshow('window', new_screen)
cv2.imshow('window2',cv2.cvtColor(original_image, cv2.COLOR_BGR2RGB))
if m1 < 0 and m2 < 0:
right()
elif m1 > 0 and m2 > 0:
left()
else:
straight()
sys.path.append("..")
conf_thres = 0.3
iou_thres = 0.2
classes = None
agnostic_nms = True
names = [
'hero', 'small_map', "monster", 'money', 'material', 'door', 'BOSS', 'box',
'options'
]
colors = [[random.randint(0, 255) for _ in range(3)]
for _ in range(len(names))]
if half:
model.half() # to FP16
while (True):
if not paused:
img0 = grab_screen(window_size)
print('loop took {} seconds'.format(time.time() - last_time))
last_time = time.time()
img0 = cv2.cvtColor(img0, cv2.COLOR_BGRA2BGR)
# Padded resize
img = letterbox(img0, new_shape=img_size)[0]
# Convert
img = img[:, :, ::-1].transpose(2, 0, 1) # BGR to RGB, to 3x416x416
img = np.ascontiguousarray(img)
img = torch.from_numpy(img).to(device).unsqueeze(0)
img = img.half() if half else img.float() # uint8 to fp16/32
img /= 255.0 # 0 - 255 to 0.0 - 1.0
t1 = time_synchronized()
model = load_model_weights('v2_nfs_ai')
print("Starting !!!")
time.sleep(1)
print("1")
time.sleep(1)
print("2")
time.sleep(1)
print("3")
time.sleep(1)
print("4")
print("GO go go go !!!!")
while (True):
current_view = grab_screen([0, 0, 800, 600])
reduced_view = cv2.resize(current_view, (shape[1], shape[0]),
interpolation=cv2.INTER_AREA)
right_view = reduced_view[:, 0:shape[1] - margin]
left_view = reduced_view[:, margin:]
cv2.imshow('frame', reduced_view)
right_view = right_view.reshape(1, right_view.shape[0],
right_view.shape[1], 1)
left_view = left_view.reshape(1, left_view.shape[0],
left_view.shape[1], 1)
instant_move = key_check()
res = model.predict({
'right_input': right_view,
def main():
last_time = time.time()
for i in list(range(4))[::-1]:
print(i + 1)
time.sleep(1)
paused = False
while (True):
if not paused:
# 800x600 windowed mode
#screen = np.array(ImageGrab.grab(bbox=(0,40,800,640)))
screen = grab_screen(region=(0, 40, 800, 640))
print('loop took {} seconds'.format(time.time() - last_time))
last_time = time.time()
screen = cv2.cvtColor(screen, cv2.COLOR_BGR2GRAY)
screen = cv2.resize(screen, (WIDTH, HEIGHT))
inp = np.reshape(screen, (1, 1, WIDTH, HEIGHT))
print(inp.shape)
#inp = [screen.reshape(WIDTH,HEIGHT,1)]
prediction = model.forward(torch.cuda.FloatTensor(inp))[0]
print(prediction)
turn_thresh = .75
fwd_thresh = 0.70
slow_tresh = 0.8
if prediction[1] > fwd_thresh:
straight()
elif prediction[0] > turn_thresh:
left()
elif prediction[2] > turn_thresh:
right()
elif prediction[3] > slow_tresh:
slow()
else:
straight()
keys = key_check()
# p pauses game and can get annoying.
if 'T' in keys:
if paused:
paused = False
time.sleep(1)
else:
paused = True
ReleaseKey(A)
ReleaseKey(W)
ReleaseKey(D)
time.sleep(1)
def main():
'''
with the z axis, your %s are out of 32,786
with the x and y, your %s are out of 16393
...so left = 16393 - (some % of 16393) ... right = 16393 + (some % of 16393)
'''
################
XYRANGE = 16393
ZRANGE = 32786
wAxisX = 16393
wAxisY = 16393
wAxisZ = 0
wAxisXRot = 16393
wAxisYRot = 16393
wAxisZRot = 0
last_time = time.time()
for i in list(range(4))[::-1]:
print(i+1)
time.sleep(1)
#how_long_since_move = 0
paused = False
mode_choice = 0
screen = grab_screen(region=(0,40,GAME_WIDTH,GAME_HEIGHT+40))
screen = cv2.cvtColor(screen, cv2.COLOR_BGR2RGB)
prev = cv2.resize(screen, (160,90))
t_minus = prev
t_now = prev
t_plus = prev
while(True):
if not paused:
screen = grab_screen(region=(0,40,GAME_WIDTH,GAME_HEIGHT+40))
screen = cv2.cvtColor(screen, cv2.COLOR_BGR2RGB)
last_time = time.time()
screen = cv2.resize(screen, (160,90))
delta_view = delta_images(t_minus, t_now, t_plus)
retval, delta_view = cv2.threshold(delta_view, 16, 255, 3)
cv2.normalize(delta_view, delta_view, 0, 255, cv2.NORM_MINMAX)
img_count_view = cv2.cvtColor(delta_view, cv2.COLOR_RGB2GRAY)
delta_count = cv2.countNonZero(img_count_view)
dst = cv2.addWeighted(screen,1.0, delta_view,0.6,0)
now=time.time()
delta_count_last = delta_count
t_minus = t_now
t_now = t_plus
t_plus = screen
t_plus = cv2.blur(t_plus,(4,4))
o_prediction = model.predict([screen.reshape(160,90,3)])[0]
# w s a d wa wd sa sd nk
prediction = np.array(o_prediction) * np.array([4.5, 0.1, 0.1, 0.1, 1.8, 1.8, 0.5, 0.5, 0.2])
## w s a d wa wd sa sd nk
joy_choices = np.array(o_prediction) * np.array([4.5, 2.0, 1.0, 1.0, 1.8, 1.8, 1.0, 1.0, 1.0])
# could in theory be a negative.
# w s sa sd nk
throttle = joy_choices[0] - joy_choices[1] - joy_choices[6] - joy_choices[7] - joy_choices[8]
# - is left.. .+ is right. (16393 + (-/+ up to 16393))
# a wa sa d wd sd
turn = (-1*joy_choices[2]) +(-1*joy_choices[4]) +(-1*joy_choices[6]) + joy_choices[3] + joy_choices[5] + joy_choices[7]
if throttle < -1 : throttle = -1
elif throttle > 1 : throttle = 1
if turn < -1 : turn = -1
elif turn > 1 : turn = 1
motion_log.append(delta_count)
motion_avg = round(mean(motion_log),3)
fps = 1 / round(time.time()-last_time, 3)
if throttle > 0:
vj.open()
joystickPosition = vj.generateJoystickPosition(wAxisZ=int(ZRANGE*throttle),wAxisX=int(XYRANGE + (turn*XYRANGE)))
vj.update(joystickPosition)
time.sleep(0.001)
vj.close()
print('FPS {}. Motion: {}. ThumbXaxis: {}. Throttle: {}. Brake: {}'.format(fps , motion_avg, int(XYRANGE + (turn*XYRANGE)), int(ZRANGE*throttle),0))
else:
vj.open()
joystickPosition = vj.generateJoystickPosition(wAxisZRot=int(-1*(ZRANGE*throttle)),wAxisX=int(XYRANGE + (turn*XYRANGE)))
vj.update(joystickPosition)
time.sleep(0.001)
vj.close()
print('FPS {}. Motion: {}. ThumbXaxis: {}. Throttle: {}. Brake: {}'.format(fps , motion_avg, int(XYRANGE + (turn*XYRANGE)), 0, int(-1*(ZRANGE*throttle))))
mode_choice = np.argmax(prediction)
if motion_avg < motion_req and len(motion_log) >= log_len:
print('WERE PROBABLY STUCK FFS, initiating some evasive maneuvers.')
# 0 = reverse straight, turn left out
# 1 = reverse straight, turn right out
# 2 = reverse left, turn right out
# 3 = reverse right, turn left out
quick_choice = random.randrange(0,4)
if quick_choice == 0:
reverse()
time.sleep(random.uniform(1,2))
forward_left()
time.sleep(random.uniform(1,2))
elif quick_choice == 1:
reverse()
time.sleep(random.uniform(1,2))
forward_right()
time.sleep(random.uniform(1,2))
elif quick_choice == 2:
reverse_left()
time.sleep(random.uniform(1,2))
forward_right()
time.sleep(random.uniform(1,2))
elif quick_choice == 3:
reverse_right()
time.sleep(random.uniform(1,2))
forward_left()
time.sleep(random.uniform(1,2))
for i in range(log_len-2):
del motion_log[0]
keys = key_check()
# p pauses game and can get annoying.
if 'T' in keys:
ultimate_release()
if paused:
paused = False
time.sleep(1)
else:
paused = True
ReleaseKey(A)
ReleaseKey(W)
ReleaseKey(D)
time.sleep(1)
def main():
last_time = time.time()
for i in list(range(5))[::-1]:
print(i + 1)
time.sleep(1)
paused = False
while (True):
if not paused:
# 800x600 windowed mode
screen = grab_screen(region=(window_size))
print('loop took {} seconds'.format(time.time() - last_time))
last_time = time.time()
screen = cv2.cvtColor(screen, cv2.COLOR_BGR2GRAY)
screen = cv2.resize(screen, (WIDTH, HEIGHT))
prediction = model.predict([screen.reshape(WIDTH, HEIGHT, 1)])[0]
print(prediction)
print("np后的预测:", np.argmax(prediction))
if np.argmax(prediction) == np.argmax(w):
print("前")
directkeys.go_forward()
elif np.argmax(prediction) == np.argmax(a):
print("左")
directkeys.go_left()
elif np.argmax(prediction) == np.argmax(s):
print("后")
directkeys.go_back()
elif np.argmax(prediction) == np.argmax(d):
print("右")
directkeys.go_right()
elif np.argmax(prediction) == np.argmax(q):
print("蓝拳")
directkeys.blue_fist()
elif np.argmax(prediction) == np.argmax(e):
print("红刀")
directkeys.red_trigger()
elif np.argmax(prediction) == np.argmax(j):
print("j键，刀平A")
directkeys.blade()
elif np.argmax(prediction) == np.argmax(k):
print("ctrl")
directkeys.ctrl()
elif np.argmax(prediction) == np.argmax(l):
print("l键，枪平a")
directkeys.gun()
elif np.argmax(prediction) == np.argmax(p):
print("shift")
directkeys.shift()
elif np.argmax(prediction) == np.argmax(jump):
print("跳跃")
directkeys.jump()
elif np.argmax(prediction) == np.argmax(none):
print("不动")
keys = key_check()
if 'Y' in keys:
if paused:
paused = False
time.sleep(1)
else:
paused = True
time.sleep(1)
break
def get_state():
state = grab_screen(region = capture_region)
state = cv2.cvtColor(state, cv2.COLOR_BGR2GRAY)
state = cv2.resize(state,(reshape_size[1],reshape_size[0]))
#state = state.reshape(reshape_size[0],reshape_size[1],1)
return state
v = 0
move = [0, 0, 0]
isLeftPressedKey = False
isRightPressedKey = False
isSpacePressedKey = False
window_upleft = None
window_bottomright = None
while (True):
keys = key_check()
if 'T' in keys or 't' in keys:
break
screen = grab_screen(region=(0, 0, 1300, 850))
screen = cv2.cvtColor(screen, cv2.COLOR_BGR2GRAY)
if window_upleft is None:
cv2.imwrite('s.png', screen)
screen = cv2.imread('s.png', 0)
marca = cv2.imread("marca.png", 0)
result = cv2.matchTemplate(screen, marca, cv2.TM_CCOEFF_NORMED)
origin = np.unravel_index(result.argmax(), result.shape)
print('origin', origin)
window_upleft = [origin[0] + 45, origin[1]]
print('window_upleft', window_upleft)
marca_fim = cv2.imread("marca_fim.png", 0)
result = cv2.matchTemplate(screen, marca_fim, cv2.TM_CCOEFF_NORMED)
position_right = np.unravel_index(result.argmax(), result.shape)
def main():
last_time = time.time()
for i in list(range(4))[::-1]:
print(i + 1)
time.sleep(1)
paused = False
mode_choice = 0
screen = grab_screen(region = (0, 40, GAME_WIDTH, GAME_HEIGHT + 40))
screen = cv2.cvtColor(screen, cv2.COLOR_BGR2RGB)
prev = cv2.resize(screen, (WIDTH,HEIGHT))
t_minus, t_now, t_plus = prev, prev, prev
while(True):
if not paused:
screen = grab_screen(region = (0, 40, GAME_WIDTH, GAME_HEIGHT + 40))
screen = cv2.cvtColor(screen, cv2.COLOR_BGR2RGB)
last_time = time.time()
screen = cv2.resize(screen, (WIDTH,HEIGHT))
delta_count_last = motion_detection(t_minus, t_now, t_plus)
t_minus = t_now
t_now = t_plus
t_plus = screen
t_plus = cv2.blur(t_plus, (4, 4))
prediction = model.predict([screen.reshape(WIDTH,HEIGHT, 3)])[0]
prediction = np.array(prediction) * np.array([4.5, 0.1, 0.1, 0.1, 1.8, 1.8, 0.5, 0.5, 0.2])
mode_choice = np.argmax(prediction)
if mode_choice == 0:
straight()
choice_picked = 'straight'
elif mode_choice == 1:
reverse()
choice_picked = 'reverse'
elif mode_choice == 2:
left()
choice_picked = 'left'
elif mode_choice == 3:
right()
choice_picked = 'right'
elif mode_choice == 4:
forward_left()
choice_picked = 'forward+left'
elif mode_choice == 5:
forward_right()
choice_picked = 'forward+right'
elif mode_choice == 6:
reverse_left()
choice_picked = 'reverse+left'
elif mode_choice == 7:
reverse_right()
choice_picked = 'reverse+right'
elif mode_choice == 8:
no_keys()
choice_picked = 'nokeys'
motion_log.append(delta_count)
motion_avg = round(mean(motion_log), 3)
print('loop took {0} seconds. Motion: {1}. Choice: {2}'.format(round(time.time() - last_time, 3), motion_avg, choice_picked))
if motion_avg < motion_req and len(motion_log) >= log_len:
print('WERE PROBABLY STUCK FFS, initiating some evasive maneuvers.')
# 0 = reverse straight, turn left out
# 1 = reverse straight, turn right out
# 2 = reverse left, turn right out
# 3 = reverse right, turn left out
quick_choice = random.randrange(0, 4)
if quick_choice == 0:
reverse()
time.sleep(random.uniform(1, 2))
forward_left()
time.sleep(random.uniform(1, 2))
elif quick_choice == 1:
reverse()
time.sleep(random.uniform(1, 2))
forward_right()
time.sleep(random.uniform(1, 2))
elif quick_choice == 2:
reverse_left()
time.sleep(random.uniform(1, 2))
forward_right()
time.sleep(random.uniform(1, 2))
elif quick_choice == 3:
reverse_right()
time.sleep(random.uniform(1, 2))
forward_left()
time.sleep(random.uniform(1, 2))
for i in range(log_len - 2):
del motion_log[0]
keys = key_check()
# p pauses game and can get annoying.
if 'T' in keys:
if paused:
paused = False
time.sleep(1)
else:
paused = True
ReleaseKey(A)
ReleaseKey(W)
ReleaseKey(D)
time.sleep(1)