def detect(self, frame):
height, width, ch = frame.shape
center = (width / 2, height / 2)
regions_of_interest, _ = self.preprocess.get_interest_regions(frame)
for roi in regions_of_interest:
utils.draw_red_box(frame, roi)
# path = self.classifier.classify(frame, regions_of_interest)
if regions_of_interest:
path = utils.get_max_area(regions_of_interest)
else:
path = None
path_shape = self.get_shape(path, self.shape_buffer)
if (path == None):
self.directions = [0, 0]
self.found = False
w, h = 0, 0
else:
x, y, w, h = path
utils.draw_blue_box(frame, path)
self.directions = utils.get_directions(center, x, y, w, h)
self.directions.append(0)
self.found = True
return (self.found, self.directions, path_shape, (w, h))
def detect(self, frame):
if frame is not None:
height, width, ch = frame.shape
center = (width / 2, height / 2)
regions_of_interest = self.preprocess.get_interest_regions(frame)
for x, y, w, h in regions_of_interest:
cv2.rectangle(frame, (x, y), (x + w, y + h),
utils.colors["red"], 2)
gate = self.classifier.classify(frame, regions_of_interest)
gate_shape = self.get_shape(gate, self.shape_buffer)
if gate_shape == self.shapes[3] or gate_shape == self.shapes[1]:
gate = None
if (gate == None):
self.directions = [0, 0]
self.found = False
gate_shape = None
w, h = 0, 0
else:
x, y, w, h = gate
cv2.rectangle(frame, (x, y), (x + w, y + h),
utils.colors["blue"], 6)
self.directions = utils.get_directions(center, x, y, w, h)
self.found = True
return (self.found, self.directions, gate_shape, (w, h))
else:
print('error no frame')
return False, None, None, None
def minimax_succ(state):
pos = state.player_positions[state.curr_player]
succ_states = []
board = state.board.copy()
board[pos] = -1
min_fruit_time = min(len(board[0]), len(board))
if state.lifetime >= 2 * min_fruit_time:
board = np.where(board >= 3, 0, board)
for d in utils.get_directions():
i = pos[0] + d[0]
j = pos[1] + d[1]
if 0 <= i < len(board) and 0 <= j < len(board[0]) and (
board[i][j] not in [-1, 1, 2]): # then move is legal
new_pos = (i, j)
fruit_score = board[new_pos]
players_score = list(state.players_score)
players_score[state.curr_player] += fruit_score
player_positions = list(state.player_positions)
player_positions[state.curr_player] = new_pos
old_board_value = board[new_pos]
board[new_pos] = (state.curr_player + 1)
yield SearchAlgos.State(board.copy(), tuple(players_score),
tuple(player_positions),
1 - state.curr_player, state.penalty, d,
state.lifetime + 1, state.initial_pos)
# reset the board: positions + scores
board[new_pos] = old_board_value
players_score[state.curr_player] -= fruit_score
def is_stuck(state:State,player_type):
for d in get_directions():
i = state.locations[player_type][X] + d[X]
j = state.locations[player_type][Y] + d[Y]
if legal_move(state.board,i,j):
return False
return True
def detect(self):
ret, frame = self.cap.read()
height, width, ch = frame.shape
center = (width / 2, height / 2)
regions_of_interest = self.preprocess.get_interest_regions(frame)
for x, y, w, h in regions_of_interest:
cv2.rectangle(frame, (x, y), (x + w, y + h), utils.colors["red"],
2)
#cv2.imshow('frame', frame)
gate = self.classifier.classify(frame, regions_of_interest)
if (gate == None):
self.directions = [0, 0]
self.found = False
else:
x, y, w, h = gate
cv2.rectangle(frame, (x, y), (x + w, y + h), utils.colors["blue"],
6)
self.directions = utils.get_directions(center, x, y, w, h)
self.found = True
self.out.write(frame)
return self.found, self.directions
def detect(self, frame):
height, width, ch = frame.shape
center = (width / 2, height / 2)
regions_of_interest = self.preprocess.get_interest_regions(frame)
for x, y, w, h in regions_of_interest:
cv2.rectangle(frame, (x, y), (x + w, y + h), utils.colors["red"],
2)
# have idea to use get_shape here so we classify only squares?
# add each to self.shape_list? - good for testing
# roi = (x, y, w, h)
# self.shape_list.add(roi) or
# gate_rois = self.classifier.classify(frame, roi) etc..
#cv2.imshow('frame', frame)
gate = self.classifier.classify(frame, regions_of_interest)
gate_shape = self.get_shape(gate, self.shape_buffer)
if (gate == None):
self.directions = [0, 0]
self.found = False
w, h = 0, 0
else:
x, y, w, h = gate
cv2.rectangle(frame, (x, y), (x + w, y + h), utils.colors["blue"],
6)
self.directions = utils.get_directions(center, x, y, w, h)
self.found = True
return (self.found, self.directions, gate_shape, (w, h))
def other_player_stuck(board, pos):
for d in utils.get_directions():
i = pos[0] + d[0]
j = pos[1] + d[1]
if 0 <= i < len(board) and 0 <= j < len(board[0]) and (board[i][j] not in [-1, 1, 2]):
return False
return True
def search_die(self, value):
candidate_dice = self.pp.get_interest_areas()
dice = [die for die in candidate_dice if self.classifier.predict(die) > .1]
for die in dice:
if self.get_dots(die) == value:
self.directions = utils.get_directions(die)
def __init__(self, game_time, penalty_score):
"""
Player initialization.
"""
self.game_time = game_time
self.penalty_score = penalty_score
self.directions = utils.get_directions() #[(1, 0), (0, 1), (-1, 0), (0, -1)]
def availables(board,loc):
steps_available = 0
for d in get_directions():
i = loc[0] + d[0]
j = loc[1] + d[1]
if 0 <= i < len(board) and 0 <= j < len(board[0]) and board[i][j] == 0: # then move is legal
steps_available += 1
return steps_available
def number_of_legal_cells_from_location(object_input, location):
steps_number = 0
for d in utils.get_directions():
i = location[0] + d[0]
j = location[1] + d[1]
if 0 <= i < len(object_input.game_board) and 0 <= j < len(object_input.game_board[0]) and object_input.game_board[i][j] not in [-1, 1, 2]:
steps_number += 1
return steps_number
def search_die(self, frame, value):
found = False
dice = self.locate_dice(frame)
for die in dice:
if self.get_dots(die) == value:
found = True
self.directions = utils.get_directions(die)
return found, self.directions
def is_goal(state):
pos = state.player_positions[state.curr_player]
for d in get_directions():
i = pos[0] + d[0]
j = pos[1] + d[1]
if 0 <= i < len(state.board) and 0 <= j < len(state.board[0]) and (
state.board[i][j] not in [-1, 1, 2]): # then move is legal
return False
return True
def calc_left_moves(board, pos):
sum_moves = 1
for direction in utils.get_directions():
i = pos[0] + direction[0]
j = pos[1] + direction[1]
if 0 <= i < len(board) and 0 <= j < len(
board[0]) and (board[i][j] not in [-1, 1, 2]):
board[i][j] = -1
sum_moves += calc_left_moves(board, (i, j))
return sum_moves
def heuristic_distance_from_fruit(state):
board = state.board
pos = state.player_positions[0]
for d in utils.get_directions():
i = pos[0] + d[0]
j = pos[1] + d[1]
if 0 <= i < len(board) and 0 <= j < len(
board[0]) and (board[i][j] not in [-1, 1, 2]):
return board[i][j] / state.penalty
return 0
def detect(self, frame):
print 'testing slots detect'
if frame is not None:
height, width, ch = frame.shape
center = (width / 2, height / 2)
regions_of_interest = self.preprocess.get_interest_regions(frame)
for x, y, w, h in regions_of_interest:
cv2.rectangle(frame, (x, y), (x + w, y + h),
utils.colors["red"], 2)
gate = self.classifier.classify(frame, regions_of_interest)
gate_shape = self.get_shape(gate, self.shape_buffer)
if gate_shape == self.shapes[3] or gate_shape == self.shapes[1]:
gate = None
if (gate == None):
self.directions = [0, 0]
self.found = False
gate_shape = None
w, h = 0, 0
else:
x, y, w, h = gate
cv2.rectangle(frame, (x, y), (x + w, y + h),
utils.colors["blue"], 6)
w_pad = w / 7
h_pad = h / 7
if self.is_direction_center:
self.directions = utils.get_directions(center, x, y, w, h)
cv2.rectangle(frame, (x + (3 * w_pad), y + (3 * h_pad)),
(x + (4 * w_pad), y + (4 * h_pad)),
utils.colors["green"], 2)
else:
if self.is_red_left:
self.directions = utils.get_directions_left(
center, x, y, w, h)
cv2.rectangle(frame,
(x + (2 * w_pad), y + (3 * h_pad)),
(x + (3 * w_pad), y + (4 * h_pad)),
utils.colors["green"], 2)
else:
self.directions = utils.get_directions_right(
center, x, y, w, h)
cv2.rectangle(frame,
(x + (4 * w_pad), y + (3 * h_pad)),
(x + (5 * w_pad), y + (4 * h_pad)),
utils.colors["green"], 2)
self.found = True
return (self.found, self.directions, gate_shape, (w, h))
else:
print('error no frame')
return False, None, None, None
def __init__(self,
board,
players_positions,
max_fruit_score,
max_fruit_time,
fruits_max_part_of_free_spaces=0.2,
animated=False,
animation_func=None):
"""Initialize the game properties with parameters.
input:
- board: 2D np.array. The initial board
- players_positions: the initial players positions
- max_fruit_score: max score for a fruit.
- max_fruit_time: max time for a fruit to be on board.
- fruits_max_part_of_free_spaces: the max part on board the fruits can take at each timestamp.
- animated: bool. Animated game (not in terminal) if true.
- animated_func: the function doing the animation.
"""
assert len(players_positions) == 2, 'Supporting 2 players only'
self.map = board
self.max_fruit_score = max_fruit_score
self.max_fruit_time = max_fruit_time
# min fruit time is set by the min dimension of the board
self.min_fruit_time = min(len(board[0]), len(board))
# in case that min_fruit_time is greater/equal to max_fruit_time -> change max_fruit_time
if self.max_fruit_time <= self.min_fruit_time:
self.max_fruit_time = self.min_fruit_time + 1
self.fruits_max_part_of_free_spaces = fruits_max_part_of_free_spaces
self.players_positions = players_positions
self.players_score = [0, 0] # init scores for each player
self.directions = utils.get_directions()
# Fruits:
fruits_dir = os.path.join(os.path.dirname(os.path.abspath(__file__)),
'fruits_imgs')
self.fruits_paths = [
os.path.join(fruits_dir, fruit_file)
for fruit_file in os.listdir(fruits_dir)
]
self.fruits_on_board = {}
self.turn = 0
self.animated = animated
self.animation_func = animation_func
if self.animated:
self.init_animation()
#TODO
self.create_fruits()
self.players_positions = [
tuple(reversed(position)) for position in self.players_positions
]
def get_legal_moves(board, location):
legal_moves = []
for d in get_directions():
i = location[0] + d[0]
j = location[1] + d[1]
# check legal move
if 0 <= i < len(board) and 0 <= j < len(
board[0]) and (board[i][j] not in [-1, 1, 2]):
legal_moves.append((i, j))
return legal_moves
def make_move(self, time_limit, players_score):
"""Make move with this Player.
input:
- time_limit: float, time limit for a single turn.
output:
- direction: tuple, specifing the Player's movement, chosen from self.directions
"""
start_time = time.time()
d = 1
reach_the_end = False
best_direction = None
chosen_state = None
time_limit = (2 * self.game_time *
float(self.player_turns - self.turns + 1)) / (
(self.player_turns + 1) * self.player_turns)
time_limit += self.spaire_time
if time_limit >= 5:
TIME_ESTIMATION = 0.9
else:
TIME_ESTIMATION = 0.85
#print(f'Time limit: {time_limit}')
while not reach_the_end: # and d < len(state.board)*len(state.board[0]):
iter_time_limit = TIME_ESTIMATION * (time_limit -
(time.time() - start_time))
#print(f'>>>Iter time: {iter_time_limit}')
state = State(get_directions(), self.board, self.locations,
self.fruits_on_board_dict, PLAYER, players_score,
self.penalty_score, self.fruits_ttl, self.turns)
try:
_, best_direction, reach_the_end, chosen_state = self.alphabeta.search(
state,
d,
True,
iter_time_limit,
alpha=float('-inf'),
beta=float('inf'))
d += 1
except Exception as e:
self.spaire_time = time_limit - (time.time() - start_time)
break
# Set new location
if best_direction == None:
best_direction = self.get_random_move()
self.set_player_location(best_direction)
self.turns += 1
return best_direction
def heuristic_num_steps(board, pos):
num_steps_available = 0
for d in utils.get_directions():
i = pos[0] + d[0]
j = pos[1] + d[1]
# check legal move
if 0 <= i < len(board) and 0 <= j < len(board[0]) and (board[i][j] not in [-1, 1, 2]):
num_steps_available += 1
if num_steps_available == 0:
return 0
return 1 / num_steps_available
def __init__(self, game_time, penalty_score):
AbstractPlayer.__init__(
self, game_time, penalty_score
) # keep the inheritance of the parent's (AbstractPlayer) __init__()
# Might need to add here usage of minimax.
self.max_fruit_turn = None
self.penalty_score = penalty_score
self.directions = utils.get_directions()
# TODO: Remember update this
self.current_turn = 0
self.board = None
self.pos = None
self.minimax_algo = AlphaBeta(self.utility, self.succ, None)
def get_moves_from_location(state, maximizing_player):
current_location = state.location if maximizing_player else state.rival_location
board = state.game_board
available_moves = []
directions = utils.get_directions()
for d in directions:
current_i = current_location[0] + d[0]
current_j = current_location[1] + d[1]
if 0 <= current_i < len(board) and 0 <= current_j < len(board[0]) and board[current_i][current_j] not in [-1, 1, 2]:
available_moves.append((d[0], d[1]))
return available_moves
def api_station_directions():
from_name = request.args.get('from')
to_name = request.args.get('to')
from_station = Station.nodes.filter(short__iexact=from_name)[0]
to_station = Station.nodes.filter(short__iexact=to_name)[0]
current_user.stats.directed += 1
current_user.save()
result = utils.get_directions(from_station, to_station)
return jsonify(result)
def state_score(board, pos):
num_steps_available = 0
for d in utils.get_directions():
i = pos[0] + d[0]
j = pos[1] + d[1]
# check legal move
if 0 <= i < len(board) and 0 <= j < len(board[0]) and (board[i][j] not in [-1, 1, 2]):
num_steps_available += 1
if num_steps_available == 0:
return -1
else:
return 4 - num_steps_available
def __init__(self, game_time, penalty_score):
AbstractPlayer.__init__(
self, game_time, penalty_score
) # keep the inheritance of the parent's (AbstractPlayer) __init__()
# TODO: initialize more fields, if needed, and the Minimax algorithm from SearchAlgos.py
# Might need to add here usage of minimax.
self.max_fruit_turn = None
self.penalty_score = penalty_score
self.directions = utils.get_directions()
# TODO: Remember update this
self.current_turn = 0
self.board = None
self.pos = None
self.minimax_algo = MiniMax(self.utility, self.succ, None)
def is_enemy_reachable(board, pos):
queue = deque([(pos, 0)])
seen = {pos}
while queue:
pos, distance = queue.popleft()
if board[pos] == 2:
return True
for d in utils.get_directions():
i = pos[0] + d[0]
j = pos[1] + d[1]
if 0 <= i < len(board) and 0 <= j < len(board[0]) and (
board[i][j] not in [-1, 1]) and (i, j) not in seen:
queue.append(((i, j), distance + 1))
seen.add((i, j))
return False
def can_I_move(board, pos):
num_steps_available = 0
for d in get_directions():
i = pos[0] + d[0]
j = pos[1] + d[1]
# check legal move
if 0 <= i < len(board) and 0 <= j < len(
board[0]) and (board[i][j] not in [-1, 1, 2]):
# print(num_steps_available, '^^^^^^^^^^^^^^^^^^^^^^^^^^^^^')
num_steps_available += 1
# print(pos, 'line 33')
if num_steps_available == 0:
return False
return True
def detect(self, frame, die_num=None):
if die_num is None:
die_num = 5
if frame is not None:
interest_regions = self.preprocessor.get_interest_regions(frame)
# die = [die for die in interest_regions if self.classifier.predict(die) > .1]
classified_rois = self.classifier.classify(frame, interest_regions)
dice = self.has_2_neighbors(classified_rois, self.shapes[3])
#TODO we need to implement a way to use self.die_1 and self.die_2
# in the detect to return the right coordinates. we can check if the die
# is touched by the sub if the frame is equal to the region of interest (744x480)
# or (640x480) for the laptop camera
# we can also use the dictionary, whichever one is easier
for x, y, w, h in classified_rois:
cv2.rectangle(frame, (x, y), (x + w, y + h),
utils.colors["red"], 2)
ht, wd, ch = frame.shape
dice_shape = self.get_shape(dice, self.shape_buffer)
# if dice_shape != self.shapes[1]:
# dice = None
if not dice:
self.found = False
dice_shape = None
self.directions = [0, 0]
w, h = 0, 0
else:
x, y, w, h = dice
# dice_shape = self.get_shape(dice, self.shape_buffer)
# dice_shape = None
cv2.rectangle(frame, (x, y), (x + w, y + h),
utils.colors["blue"], 6)
self.directions = utils.get_directions((wd / 2, ht / 2), x, y,
w, h)
self.found = True
#found, direction, shape, width, heightk
# return (self.found, self.directions, None, (0, 0))
return (self.found, self.directions, dice_shape, (w, h))
else:
print('error no frame')
return False, None, None, None
def make_move(self, time_limit, players_score):
"""Make move with this Player.
input:
- time_limit: float, time limit for a single turn.
output:
- direction: tuple, specifing the Player's movement, chosen from self.directions
"""
start_time = time.time()
d = 1
# Make the initial state:
reach_the_end = False
best_direction = None
chosen_state = None
if time_limit >= 5:
TIME_ESTIMATION = 0.9
else:
TIME_ESTIMATION = 0.85
while not reach_the_end:
iter_time_limit = TIME_ESTIMATION * (time_limit -
(time.time() - start_time))
state = State(get_directions(), self.board, self.locations,
self.fruits_on_board_dict, PLAYER, players_score,
self.penalty_score, self.fruits_ttl, self.turns)
try:
_, best_direction, reach_the_end, chosen_state = self.minimax.search(
state, d, True, iter_time_limit)
d += 1
except Exception as e:
break
# Set new location
if best_direction == None:
best_direction = self.get_random_move()
self.set_player_location(best_direction)
self.turns += 1
return best_direction
def respond(request):
"""Respond to an HTTP request
Parameters
----------
request
Returns
-------
"""
text = request.form.get("text", None)
channel = request.form.get("channel_id")
start, end = text.split(" to ")
path_nodes, path_edges = get_directions(start, end)
plot_path(path_nodes, path_edges)
plt.savefig("directions.png", format="png", dpi=200)
client = slack.WebClient(token=os.environ['SLACK_API_TOKEN'])
response = client.files_upload(channels=channel, file="directions.png")