def forward(self, input, is_train=False):
_input = noisy(self.vocab, input, *
self.args.noise) if is_train else input
mu, logvar = self.encode(_input)
z = reparameterize(mu, logvar)
logits, _ = self.decode(z, input)
return mu, logvar, z, logits
def forward(self, input):
if self.training:
_input = noisy(self.padidx, input, 0.3, 0, 0, 0)
else:
_input = input
#mu, logvar = self.encode(_input)
h = self.encode(_input)
return h #mu, logvar
def eval_genomes(genomes, config):
for genome_id,genome in genomes:
ob = env.reset()
inx, iny, inc = env.observation_space.shape
inx = int(inx/8)
iny = int(iny/8)
net = neat.nn.recurrent.RecurrentNetwork.create(genome, config)
fitness_current = 0
frame = 0
counter = 0
done = False
while not done:
frame += 1
factor = 0.5
ob = np.uint8(noise.noisy(ob, factor))
ob = cv2.resize(ob, (inx,iny))
ob = cv2.cvtColor(ob, cv2.COLOR_BGR2GRAY)
imageArr = np.ndarray.flatten(ob)
nnOutput = net.activate(imageArr)
numerical_input = nnOutput.index(max(nnOutput))
ob, rew, done, info = env.step(numerical_input)
fitness_current += rew
if(rew > 0):
counter = 0
else:
counter += 1
env.render()
if(done or counter == 250):
done = True
print(genome_id, fitness_current)
genome.fitness = fitness_current
def eval_genomes(genomes, config):
for genome_id, genome in genomes:
observation = env.reset()
inputx, inputy, inputColour = env.observation_space.shape
inputx = int(inputx / 8)
inputy = int(inputy / 8)
net = neat.nn.recurrent.RecurrentNetwork.create(genome, config)
fitness_current = 0
frame = 0
counter = 0
game_done = False
while not game_done:
frame += 1
factor = 0.5
observation = np.uint8(noise.noisy(observation, factor))
observation = cv2.resize(observation, (inputx, inputy))
#observation = cv2.cvtColor(observation, cv2.COLOR_RGB2GRAY)
imagearray = np.ndarray.flatten(observation)
nnOutput = net.activate(imagearray)
numerical_input = nnOutput.index(max(nnOutput))
observation, reward, game_done, info = env.step(numerical_input)
fitness_current += reward
if reward > 0:
counter = 0
else:
counter += 1
env.render()
if game_done or counter == 300:
game_done = True
print(genome_id, fitness_current, counter)
genome.fitness = fitness_current
def imageCapture(self):
monitor = {'left': 0, 'top': 0, 'width': self.w, 'height': self.h}
with mss.mss() as sct:
while True:
sct_img = sct.grab(monitor)
img_np = np.array(sct_img)
if self.noise:
img_np = noise.noisy(img_np, self.noiseType)
img_np = cv2.normalize(img_np, None, 0, 255,
cv2.NORM_MINMAX).astype('uint8')
gray_frame = cv2.cvtColor(img_np, cv2.COLOR_BGR2GRAY)
threash_triangle = threshold_triangle(gray_frame)
binary_triangle = gray_frame > threash_triangle
binary_triangle = img_as_ubyte(binary_triangle)
cropped_bw_frame = binary_triangle[
int(7 * (self.h / 13.0)):int(self.h - (self.h / 5)),
int(self.w / 5):int(self.w - (self.w / 5))]
resized_bw_frame = cv2.resize(cropped_bw_frame,
(int(160), int(90)),
interpolation=cv2.INTER_AREA)
self.bw_frame = cv2.bitwise_not(resized_bw_frame)
self.rgb_frame = cv2.cvtColor(img_np, cv2.COLOR_BGR2RGB)
_ = cv2.waitKey(1)
import noise
import numpy as np
import cv2
img = cv2.imread('buried.jpg')
img = np.uint8(noise.noisy(img, 0.5))
cv2.imshow("image", img)
cv2.waitKey(0)
cv2.desAllWindows()
out1 = VideoWriter('s&p_2.mp4', fourcc, 60.0, (w4, h4))
in2 = VideoWriter('gauss_1.mp4', fourcc, 60.0, (w1, h1))
out2 = VideoWriter('gauss_2.mp4', fourcc, 60.0, (w4, h4))
in3 = VideoWriter('poisson_1.mp4', fourcc, 60.0, (w1, h1))
out3 = VideoWriter('poisson_2.mp4', fourcc, 60.0, (w4, h4))
in4 = VideoWriter('speckle_1.mp4', fourcc, 60.0, (w1, h1))
out4 = VideoWriter('speckle_2.mp4', fourcc, 60.0, (w4, h4))
bw1 = list()
bw2 = list()
bw3 = list()
bw4 = list()
for i in tqdm(range(int(cap.get(7)))):
ret, frame = cap.read()
if ret:
n1 = noise.noisy(frame, "s&p")
n1 = cv2.normalize(n1, None, 0, 255, cv2.NORM_MINMAX).astype('uint8')
in1.write(n1)
gray_frame = cv2.cvtColor(n1, cv2.COLOR_BGR2GRAY)
threash_triangle = threshold_triangle(gray_frame)
binary_triangle = gray_frame > threash_triangle
binary_triangle = img_as_ubyte(binary_triangle)
cropped_bw_frame = binary_triangle[int(7 * (h / 13.0)):int(h -
(h / 5)),
int(w / 5):int(w - (w / 5))]
resized_bw_frame = cv2.resize(cropped_bw_frame, (int(160), int(90)),
interpolation=cv2.INTER_AREA)
bw_frame = cv2.bitwise_not(resized_bw_frame)
bw1.append(bw_frame)
if i > 20:
difference = cv2.bitwise_not(bw1[i - 20] - bw_frame)