def SharpEnchance(Img, factor):
enh_sha = ImageEnhance.Sharpness(Img)
image_sharped = enh_sha.enhance(factor)
return image_sharped
def main():
count = 0
#parameter
wsize = 1024 # double the resolution
Gamma = 0.97
Phi = 200
Epsilon = 0.1
k = 2
Sigma = 1.5
max_length = 20
min_length = 10
max_dif = 30
n_point = 50
dir = 3
input_paths = glob.glob(in_dir + '/*.jpg')
input_paths += (glob.glob(in_dir + '/*.jpeg'))
input_paths += (glob.glob(in_dir + '/*.png'))
for files1 in input_paths:
filepath, filename = os.path.split(files1)
im = Image.open(files1).convert('L')
im = array(ImageEnhance.Sharpness(im).enhance(3.0))
im2 = filters.gaussian_filter(im, Sigma)
im3 = filters.gaussian_filter(im, Sigma * k)
differencedIm2 = im2 - (Gamma * im3)
(x, y) = shape(im2)
for i in range(x):
for j in range(y):
if differencedIm2[i, j] < Epsilon:
differencedIm2[i, j] = 1
else:
differencedIm2[i, j] = 250 + tanh(Phi *
(differencedIm2[i, j]))
gray_pic = differencedIm2.astype(np.uint8)
color_pic = Image.open(files1)
real = np.atleast_2d(color_pic)
if real.ndim == 3:
w, h, c = real.shape
if c == 3:
image = color_pic.filter(MyGaussianBlur(radius=5))
mat = np.atleast_2d(image)
if gray_pic.ndim == 2:
gray_pic = np.expand_dims(gray_pic, 2)
gray_pic = np.tile(gray_pic, [1, 1, 3])
# for i in range(n_point):
# length = 0
# while length < min_length:
# ws, hs, wt, ht, length = gen_color_line(mat, dir,max_length,max_dif)
# gray_pic[ws:wt, hs:ht, :] = mat[ws:wt, hs:ht, :]
gray_pic = np.append(real, gray_pic, axis=1)
final_img = Image.fromarray(gray_pic)
im = final_img
w, h = im.size
hsize = int(h * wsize / float(w))
if hsize * 2 > wsize: # crop to three
im = im.resize((wsize, hsize))
bounds1 = (0, 0, wsize, int(wsize / 2))
cropImg1 = im.crop(bounds1)
# cropImg1.show()
cropImg1.save(os.path.join(out_dir, 'u' + filename))
bounds2 = (0, hsize - int(wsize / 2), wsize, hsize)
cropImg2 = im.crop(bounds2)
# cropImg.show()
cropImg2.save(os.path.join(out_dir, 'd' + filename))
else:
im = im.resize((wsize, (wsize // 2)))
im.save(os.path.join(out_dir, 't' + filename))
count += 1
print('done!' + str(count))
def transform(self, image, factor):
enhancer = ImageEnhance.Sharpness(Image.fromarray(image))
return np.array(enhancer.enhance(factor))
def __call__(self, img):
alpha = random.uniform(1 - self.var, 1 + self.var)
return ie.Sharpness(img).enhance(alpha)
async def deepfry(img: Image,
*,
token: str = None,
url_base: str = 'westcentralus',
session: aiohttp.ClientSession = None,
type=DeepfryTypes.RED) -> Image:
"""
Deepfry an image.
img: PIL.Image - Image to deepfry.
[token]: str - Token to use for Microsoft facial recognition API. If this is not supplied, lens flares will not be added.
[url_base]: str = 'westcentralus' - API base to use. Only needed if your key's region is not `westcentralus`.
[session]: aiohttp.ClientSession - Optional session to use with API requests. If provided, may provide a bit more speed.
Returns: PIL.Image - Deepfried image.
"""
img = img.copy().convert('RGB')
if type not in DeepfryTypes:
raise ValueError(
f'Unknown deepfry type "{type}", expected a value from deeppyer.DeepfryTypes'
)
if token:
req_url = f'https://{url_base}.api.cognitive.microsoft.com/face/v1.0/detect?returnFaceId=false&returnFaceLandmarks=true' # WHY THE F**K IS THIS SO LONG
headers = {
'Content-Type': 'application/octet-stream',
'Ocp-Apim-Subscription-Key': token,
'User-Agent': 'DeepPyer/1.0'
}
b = BytesIO()
img.save(b, 'jpeg')
b.seek(0)
if session:
async with session.post(req_url, headers=headers,
data=b.read()) as r:
face_data = await r.json()
else:
async with aiohttp.ClientSession() as s, s.post(
req_url, headers=headers, data=b.read()) as r:
face_data = await r.json()
if 'error' in face_data:
err = face_data['error']
code = err.get('code', err.get('statusCode'))
msg = err['message']
raise Exception(
f'Error with Microsoft Face Recognition API\n{code}: {msg}')
if face_data:
landmarks = face_data[0]['faceLandmarks']
# Get size and positions of eyes, and generate sizes for the flares
eye_left_width = math.ceil(landmarks['eyeLeftInner']['x'] -
landmarks['eyeLeftOuter']['x'])
eye_left_height = math.ceil(landmarks['eyeLeftBottom']['y'] -
landmarks['eyeLeftTop']['y'])
eye_left_corner = (landmarks['eyeLeftOuter']['x'],
landmarks['eyeLeftTop']['y'])
flare_left_size = eye_left_height if eye_left_height > eye_left_width else eye_left_width
flare_left_size *= 4
eye_left_corner = tuple(
math.floor(x - flare_left_size / 2.5 + 5)
for x in eye_left_corner)
eye_right_width = math.ceil(landmarks['eyeRightOuter']['x'] -
landmarks['eyeRightInner']['x'])
eye_right_height = math.ceil(landmarks['eyeRightBottom']['y'] -
landmarks['eyeRightTop']['y'])
eye_right_corner = (landmarks['eyeRightInner']['x'],
landmarks['eyeRightTop']['y'])
flare_right_size = eye_right_height if eye_right_height > eye_right_width else eye_right_width
flare_right_size *= 4
eye_right_corner = tuple(
math.floor(x - flare_right_size / 2.5 + 5)
for x in eye_right_corner)
# Crush image to hell and back
img = img.convert('RGB')
width, height = img.width, img.height
img = img.resize((int(width**.75), int(height**.75)),
resample=Image.LANCZOS)
img = img.resize((int(width**.88), int(height**.88)),
resample=Image.BILINEAR)
img = img.resize((int(width**.9), int(height**.9)), resample=Image.BICUBIC)
img = img.resize((width, height), resample=Image.BICUBIC)
img = ImageOps.posterize(img, 4)
# Generate red and yellow overlay for classic deepfry effect
r = img.split()[0]
r = ImageEnhance.Contrast(r).enhance(2.0)
r = ImageEnhance.Brightness(r).enhance(1.5)
if type == DeepfryTypes.RED:
r = ImageOps.colorize(r, Colours.RED, Colours.YELLOW)
elif type == DeepfryTypes.BLUE:
r = ImageOps.colorize(r, Colours.BLUE, Colours.WHITE)
# Overlay red and yellow onto main image and sharpen the hell out of it
img = Image.blend(img, r, 0.75)
img = ImageEnhance.Sharpness(img).enhance(100.0)
if token and face_data:
# Copy and resize flares
flare = Image.open('./deeppyer/flare.png')
flare_left = flare.copy().resize((flare_left_size, ) * 2,
resample=Image.BILINEAR)
flare_right = flare.copy().resize((flare_right_size, ) * 2,
resample=Image.BILINEAR)
del flare
img.paste(flare_left, eye_left_corner, flare_left)
img.paste(flare_right, eye_right_corner, flare_right)
return img
def posterize_func(img, bboxes, bits):
out = np.bitwise_and(img, np.uint8(255 << (8 - bits)))
return out, bboxes
if __name__ == '__main__':
import PIL
from PIL import Image, ImageEnhance, ImageOps
import time
pth = './pic.jpg'
impil = Image.open(pth)
imcv = cv2.imread(pth)
out_cv = sharpness_func(imcv, 0.3).astype(np.float32)
sharp_pil = ImageEnhance.Sharpness(impil)
out_pil = np.array(sharp_pil.enhance(0.3))[:, :, ::-1].astype(np.float32)
print('sharpness')
print(np.sum(np.abs(out_pil - out_cv)))
print(np.max(np.abs(out_pil - out_cv)))
print(np.min(np.abs(out_pil - out_cv)))
pth = './pic.jpg'
impil = Image.open(pth)
imcv = cv2.imread(pth)
t1 = time.time()
n_test = 100
for i in range(n_test):
out_cv = autocontrast_func(imcv, 20).astype(np.float32)
t2 = time.time()
for i in range(n_test):
def sharpness(image,val):
_image = ImageEnhance.Sharpness(image)
return _image.enhance(val)
circle_status = 0
else:
circle_status = 1
if analyze_pixel(capture_bar) == capture_bar_found:
capture_bar_status = 1
else:
capture_bar_status = 0
return [circle_status, capture_bar_status, artist_box_status]
if __name__ == "__main__":
print("Simple: ", read_image("test.png"))
print()
sharp = ImageEnhance.Sharpness(Image.open("test.png"))
sharp = sharp.enhance(0.5)
sharp.save("test_sharpness.png")
print("Sharpness: ", read_image("test_sharpness.png"))
print()
bright = ImageEnhance.Brightness(Image.open("test.png"))
bright = bright.enhance(0.5)
bright.save("test_brightness.png")
print("Brightness: ", read_image("test_brightness.png"))
print()
contr = ImageEnhance.Contrast(Image.open("test.png"))
contr = contr.enhance(0.5)
contr.save("test_contrast.png")
print("Contrast: ", read_image("test_contrast.png"))
print()
color = ImageEnhance.Color(Image.open("test.png"))
def imageInflated(_inImportPath, _inExportPath):
files = os.listdir(_inImportPath)
imgNum = 0
for file in files:
index = re.search('.png',file)
if index:
imgNum += 1
print(imgNum)
for i in range(imgNum):
importImagePath = os.path.join(_inImportPath,str(i+1)+'.png')
im_0 = Image.open(importImagePath)
saveImage(im_0,_inExportPath)
im_1 = ImageOps.mirror(im_0)
saveImage(im_1,_inExportPath)
im_2 = im_0.point(lambda x:x * 1.3)
saveImage(im_2,_inExportPath)
im_3 = im_1.point(lambda x:x * 1.3)
saveImage(im_3,_inExportPath)
im_4 = im_0.point(lambda x:x * 0.7)
saveImage(im_4,_inExportPath)
im_5 = im_1.point(lambda x:x * 0.7)
saveImage(im_5,_inExportPath)
im_6 = changeColorRate(im_0,1.2,1,1)
saveImage(im_6,_inExportPath)
im_8 = changeColorRate(im_1,1.2,1,1)
saveImage(im_8,_inExportPath)
im_7 = changeColorRate(im_0,0.8,1,1)
saveImage(im_7,_inExportPath)
im_9 = changeColorRate(im_1,0.8,1,1)
saveImage(im_9,_inExportPath)
im_10 = changeColorRate(im_0,1,1.2,1)
saveImage(im_10,_inExportPath)
im_12 = changeColorRate(im_1,1,1.2,1)
saveImage(im_12,_inExportPath)
im_11 = changeColorRate(im_0,1,0.8,1)
saveImage(im_11,_inExportPath)
im_13 = changeColorRate(im_1,1,0.8,1)
saveImage(im_13,_inExportPath)
im_14 = changeColorRate(im_0,1,1,1.2)
saveImage(im_14,_inExportPath)
im_16 = changeColorRate(im_1,1,1,1.2)
saveImage(im_16,_inExportPath)
im_15 = changeColorRate(im_0,1,1,0.8)
saveImage(im_15,_inExportPath)
im_17 = changeColorRate(im_1,1,1,0.8)
saveImage(im_17,_inExportPath)
iec_1 = ImageEnhance.Contrast(im_0)
im_18 = iec_1.enhance(1.3)
saveImage(im_18,_inExportPath)
iec_2 = ImageEnhance.Contrast(im_1)
im_19 = iec_2.enhance(1.3)
saveImage(im_19,_inExportPath)
iec_3 = ImageEnhance.Contrast(im_0)
im_20 = iec_3.enhance(0.7)
saveImage(im_20,_inExportPath)
iec_4 = ImageEnhance.Contrast(im_1)
im_21 = iec_4.enhance(0.7)
saveImage(im_21,_inExportPath)
iec_5 = ImageEnhance.Color(im_0)
im_22 = iec_5.enhance(1.3)
saveImage(im_22,_inExportPath)
iec_6 = ImageEnhance.Color(im_1)
im_23 = iec_6.enhance(1.3)
saveImage(im_23,_inExportPath)
iec_7 = ImageEnhance.Color(im_0)
im_24 = iec_7.enhance(0.7)
saveImage(im_24,_inExportPath)
iec_8 = ImageEnhance.Color(im_1)
im_25 = iec_8.enhance(0.7)
saveImage(im_25,_inExportPath)
iec_9 = ImageEnhance.Sharpness(im_0)
im_26 = iec_9.enhance(2)
saveImage(im_26,_inExportPath)
iec_10 = ImageEnhance.Sharpness(im_1)
im_27 = iec_10.enhance(2)
saveImage(im_27,_inExportPath)
iec_11 = ImageEnhance.Sharpness(im_0)
im_28 = iec_11.enhance(0.3)
saveImage(im_28,_inExportPath)
iec_12 = ImageEnhance.Sharpness(im_1)
im_29 = iec_12.enhance(0.3)
saveImage(im_29,_inExportPath)
def transform_image(self, img, shape, trials=20):
# Skew
################
# w, h = img.size
# skew = random.random()*2 - 1
# xshift = abs(skew) * w
# new_width = w + int(round(xshift))
# img = img.transform(
# (new_width, h),
# Image.AFFINE,
# (1, skew, -xshift if skew > 0 else 0, 0, 1, 0),
# Image.BICUBIC
# )
# Rotate
################
theta = random.randint(-180, 180)
img = img.rotate(theta)
# Random Crop
#################
w, h = img.size
rand_points = []
c = 0
while c < trials:
try:
# rand top left corner
rand_x, rand_y = random.randint(0, int(w / 3)), random.randint(
0, int(h / 3))
# rand side length greater than half shape dimensions
rand_side = random.randint(min(shape[:2]),
min((w - rand_x), (h - rand_y)) - 1)
# generates the set of corners for random crop
rand_points = [(rand_x, rand_y), (rand_x + rand_side, rand_y),
(rand_x, rand_y + rand_side),
(rand_x + rand_side, rand_y + rand_side)]
# if all corners are part of image => move on
for x, y in rand_points:
if img.getpixel((x, y)) == (0, 0, 0):
rand_points.remove((x, y))
if len(rand_points) == 4:
break
# Try again
c += 1
# If exception occurred => try again
except:
c += 1
# If couldn't generate image in n trials => return None
if c >= trials:
return None
# Sharpening, Brightness, and Contrast
#################
sharpener = ImageEnhance.Sharpness(img)
img = sharpener.enhance(1 + (random.random() - .5) / 2)
brightener = ImageEnhance.Brightness(img)
img = brightener.enhance(1 + (random.random() - .5) / 2)
contraster = ImageEnhance.Contrast(img)
img = contraster.enhance(1 + (random.random() - .5) / 2)
# Final Resize
##################
box = (rand_x, rand_y, rand_x + rand_side, rand_y + rand_side)
img = img.resize((shape[0], shape[1]), Image.LANCZOS, box)
# img.show()
return img
def apply(self, image):
return ImageEnhance.Sharpness(image).enhance(self.value)
def sharpnessPhoto(self, level):
'''Регулировка резкости изображения'''
self.objectMedia = ImageEnhance.Sharpness(
self.objectMedia).enhance(level)
self.addHistoryChange()
def img_prep(img):
enh_bri = ImageEnhance.Brightness(img).enhance(1.05)
enh_col = ImageEnhance.Color(enh_bri).enhance(1.6)
enh_con = ImageEnhance.Contrast(enh_col).enhance(1.8)
new_img = ImageEnhance.Sharpness(enh_con).enhance(2.5)
return new_img
import pandas as pd
from PIL import Image
from PIL import ImageEnhance
im = Image.open("test.png")
double_size = (im.size[0] * 2, im.size[1] * 2)
larger_im = im.resize(double_size)
larger_im = ImageEnhance.Sharpness(larger_im).enhance(2.0)
larger_im.save("enhance.png")
with open('tableResult6.txt', 'r') as file:
a = file.read()
txt_vals = [b.split('\t') for b in a.split('\n') if b]
di_rows = []
course = []
reoc = []
start = []
end = []
DIreoc = []
DIstart = []
DIend = []
FIdate = []
FIstart = []
FIend = []
courseReach = 0
# well I need this
def hasNumber(inputString):
return any(char.isdigit() for char in inputString)
def ContrastivePredictiveCodingAugmentations(img):
# We use transformations as traceable
# https://arxiv.org/pdf/1805.09501.pdf
pool = [
transforms.RandomRotation( # Rotation
30,
resample=False,
expand=False,
center=None,
fill=None),
transforms.RandomAffine( # Shearing
0,
translate=None,
scale=None,
shear=30,
resample=False,
fillcolor=0),
transforms.RandomAffine( # Translate
0,
translate=(0.3, 0.3),
scale=None,
shear=None,
resample=False,
fillcolor=0),
transforms.Lambda(lambda x: imo.autocontrast(x)), # Autocontrast
transforms.Lambda(lambda x: imo.invert(x)), # Invert
transforms.Lambda(lambda x: imo.equalize(x)), # Equalize
transforms.Lambda(lambda x: imo.solarize(x)), # Solarize
transforms.Lambda(lambda x: imo.posterize(
x, bits=int(np.random.randint(4, 8) + 1))), # Posterize
transforms.Lambda(
lambda x: ime.Color(x).enhance(np.random.uniform())), # Color
transforms.Lambda(lambda x: ime.Brightness(x).enhance(
np.random.uniform())), # Brightness
transforms.Lambda(lambda x: ime.Contrast(x).enhance(np.random.uniform(
))), # Contrast
transforms.Lambda(lambda x: ime.Sharpness(x).enhance(np.random.uniform(
))), # Sharpness
transforms.Compose( # Set black
[
transforms.ToTensor(),
transforms.RandomErasing(1.0),
transforms.ToPILImage()
])
]
# 1.
t1 = transforms.RandomChoice(pool)
t2 = transforms.RandomChoice(pool)
t3 = transforms.RandomChoice(pool)
img = t3(t2(t1(img)))
# https://www.nature.com/articles/s41591-018-0107-6
# 2. Only elastic def, no shearing as this is part of pool as well as hist changes
if np.random.uniform() < 0.2:
img = elastic_transform(img, sigma=10)
# 3. In pool
# 4.
if np.random.uniform() < 0.25:
img = transforms.functional.to_grayscale(img, num_output_channels=3)
return img
def __getitem__(self, idx):
if self.target_network is None:
self.target_network = get_model(self.target_network_name,
gpus=[0],
num_classes=self.dataset.num_class,
train_aug=self.target_aug).eval()
img_orig, lb = self.dataset[idx]
n_img, n_lb, n_losses, n_corrects = [], [], [], []
for _ in range(self.num_sample):
tta_rotate_default = random.choice(
self.transform_r) if random.random() < self.d_tta_prob else 0.0
tta_zoom_default = random.choice(
self.transform_zs
) if random.random() < self.d_tta_prob else 1.0
tta_bright_default = random.choice(
self.transform_b) if random.random() < self.d_tta_prob else 1.0
for t_flip in self.transform_flip:
img_new = img_orig.copy()
corrupt_name = corrupt_op = corrupt_lv = None
if self.do_random_corrupt and random.random(
) < self.d_tta_prob: # TODO
corrupt_op = random.choice(corruption_tuple)
corrupt_lv = random.choice([1, 2, 3, 4, 5])
elif isinstance(self.corrupt_list, list) and random.random(
) < self.d_tta_prob: # TODO : Partial Corruptions
corrupt_name = random.choice(self.corrupt_list)
corrupt_op = corruption_dict[corrupt_name]
corrupt_lv = random.choice([1, 2, 3, 4, 5])
# corrupt_lv = random.choice([3, 4, 5])
if corrupt_op is not None:
img_np = corrupt_op(img_new, severity=corrupt_lv)
if isinstance(img_np, np.ndarray):
img_np = img_np.astype(np.uint8)
img_new = Image.fromarray(img_np)
elif isinstance(img_np, PIL.Image.Image):
img_new = img_np
else:
raise Exception(type(img_np))
if t_flip:
img_new = torchvision.transforms.functional.hflip(img_new)
mirror_expansion_factor = 3
try:
img_mirror = mirror_expansion(img_new)
except Exception as e:
print(corrupt_op, corrupt_lv)
print(e)
print(type(img_new))
print(img_new.size)
raise e
img_new = img_mirror.copy()
if tta_rotate_default != 0:
img_new = torchvision.transforms.functional.rotate(
img_new,
tta_rotate_default,
expand=False,
resample=PIL.Image.BICUBIC)
assert tta_bright_default > 0
if tta_bright_default != 1.0:
img_new = torchvision.transforms.functional.adjust_brightness(
img_new, tta_bright_default)
new_resize = int((self.target_size + self.padding) *
mirror_expansion_factor * tta_zoom_default)
assert 0.5 < tta_zoom_default < 1.5
if tta_zoom_default != 1.0:
img_new = torchvision.transforms.functional.resize(
img_new, new_resize, interpolation=PIL.Image.BICUBIC)
imgs_pil = []
for tta_action in tta_actions:
tta_rotate, tta_brightness, tta_zoom, tta_contrast, tta_color, tta_blur, tta_att, _ = decode(
tta_action)
if tta_rotate != 0:
img_rotate = torchvision.transforms.functional.rotate(
img_new,
tta_rotate,
expand=False,
resample=PIL.Image.BICUBIC)
else:
img_rotate = img_new.copy()
if tta_brightness != 1.0:
img_bright = torchvision.transforms.functional.adjust_brightness(
img_rotate, tta_brightness)
else:
img_bright = img_rotate.copy()
if tta_zoom != 1.0:
resize = int(new_resize * tta_zoom)
img_zoom = torchvision.transforms.functional.resize(
img_bright,
resize,
interpolation=PIL.Image.BICUBIC)
assert img_zoom.width > 32, (img_zoom.size,
img_bright.size)
else:
img_zoom = img_bright.copy()
if tta_contrast > 0.0:
img_zoom = PIL.ImageOps.autocontrast(img_zoom)
assert img_zoom.width > 32, ('autocont', img_zoom.size,
img_bright.size, img_new.size)
if tta_color != 1.0:
img_zoom = PIL.ImageEnhance.Color(img_zoom).enhance(
tta_color)
assert img_zoom.width > 32, ('color', img_zoom.size,
img_bright.size, img_new.size)
if tta_blur != 1.0:
img_zoom = ImageEnhance.Sharpness(img_zoom).enhance(
tta_blur)
assert img_zoom.width > 32, ('blur', img_zoom.size,
img_bright.size, img_new.size)
w, h = img_zoom.size
att_padding = self.padding if self.padding else 0
pw, ph = max(0, self.target_size + att_padding - w), max(
0, self.target_size + att_padding - h)
pw1, ph1 = pw // 2, ph // 2
pw2, ph2 = pw - pw1, ph - ph1
if pw1 > 0 or ph1 > 0 or pw2 > 0 or ph2 > 0:
img_pad = torchvision.transforms.functional.pad(
img_zoom, (pw1, ph1, pw2, ph2),
random.randint(0, 255), 'reflect')
else:
img_pad = img_zoom
# img = torchvision.transforms.functional.center_crop(img_zoom, (self.target_size, self.target_size))
crop_width = crop_height = self.target_size
# print(tta_action, 'orig.size=', img_orig.size, 'zoom.size=', img_zoom.size, 'img_pad.size', img_pad.size, 'target_size=', self.target_size, 'padding=', self.padding)
if tta_att == 0:
img = pil_center_crop(img_pad, self.target_size)
elif tta_att == 1:
img_pad = pil_center_crop(
img_pad, int(self.target_size + att_padding))
img = img_pad.crop((0, 0, crop_width, crop_height))
elif tta_att == 2:
img_pad = pil_center_crop(
img_pad, int(self.target_size + att_padding))
image_width, image_height = img_pad.size
img = img_pad.crop((image_width - crop_width, 0,
image_width, crop_height))
elif tta_att == 3:
img_pad = pil_center_crop(
img_pad, int(self.target_size + att_padding))
image_width, image_height = img_pad.size
img = img_pad.crop((0, image_height - crop_height,
crop_width, image_height))
elif tta_att == 4:
img_pad = pil_center_crop(
img_pad, int(self.target_size + att_padding))
image_width, image_height = img_pad.size
img = img_pad.crop((image_width - crop_width,
image_height - crop_height,
image_width, image_height))
else:
raise Exception
imgs_pil.append(img)
self.img_pils = imgs_pil
imgs = []
for img in imgs_pil:
img = torchvision.transforms.functional.to_tensor(img)
img = torchvision.transforms.functional.normalize(
img,
mean=[0.485, 0.456, 0.406],
std=[0.229, 0.224, 0.225])
imgs.append(img)
imgs = torch.stack(imgs).cuda()
assert len(imgs) == tta_num
with torch.no_grad():
preds = self.target_network(imgs)
corrects = (torch.argmax(
preds,
dim=1).squeeze() == lb).detach().cpu().int().float()
lbs = torch.tensor([lb] * tta_num).squeeze().cuda()
# taus = torch.FloatTensor(tta_taus).detach()
losses = torch.nn.functional.cross_entropy(
preds, lbs, reduction='none').detach().cpu()
del preds
if self.target_size > 32: # TODO
torch.cuda.empty_cache()
w, h = img_new.size
pw, ph = max(0, self.target_size + self.padding - w), max(
0, self.target_size + self.padding - h)
pw1, ph1 = pw // 2, ph // 2
pw2, ph2 = pw - pw1, ph - ph1
if pw1 > 0 or ph1 > 0 or pw2 > 0 or ph2 > 0:
img_new = torchvision.transforms.functional.pad(
img_new, (pw1, ph1, pw2, ph2), random.randint(0, 255),
'reflect')
if img_new.size[0] >= self.target_size or img_new.size[
1] >= self.target_size:
# img_new = torchvision.transforms.functional.center_crop(img_new, self.target_size)
img_new = pil_center_crop(img_new, self.target_size)
self.orig_img_pil = img_new
img_new = cutout(img_new,
cutsize=self.cutout * mirror_expansion_factor)
if self.is_test:
return img_mirror, imgs_pil, img_new, losses, corrects
img_new = torchvision.transforms.functional.resize(
img_new, self.l2t_size,
interpolation=PIL.Image.BICUBIC) # TODO
img_new = torchvision.transforms.functional.to_tensor(img_new)
img_new = torchvision.transforms.functional.normalize(
img_new,
mean=[0.485, 0.456, 0.406],
std=[0.229, 0.224, 0.225]).cpu()
n_img.append(img_new)
n_lb.append(lb)
n_losses.append(losses)
n_corrects.append(corrects)
return torch.stack(n_img), torch.Tensor(n_lb), torch.stack(
n_losses), torch.stack(n_corrects)
def enhance():
img = Image.open("./upsample/edges_001_qp30.jpg")
# enhancer1 = img.filter(ImageFilter.EDGE_ENHANCE_MORE)
enhancer1 = ImageEnhance.Sharpness(img).enhance(5.0)
enhancer1.save("test1.jpg")
def main_loop(cap):
inverted = False
autocontrast = False
black_on_yellow = False
freeze = False
magnification = 1.0
brightness = 1.0
contrast = 1.0
sharpness = 1.0
show_params = False
take_new_frame = False
UVC_functions.set_autofocus(True)
autofocus = True
focusval = 0
rotate180 = False
# default values for grabbed-frame height and width:
cam_width = 1920
cam_height = 1080
crop_width = 100
crop_height = 100
config = ConfigParser.ConfigParser()
cfg_read = config.read("EVidMag.cfg")
print(cfg_read)
if cfg_read == []:
print("Unable to read configuration file VidMag.cfg!")
else:
user_options = []
try:
user_options = ConfigSectionMap(config, "USER")
except:
print("Failed to read section [USER] from configuration file!")
print("User options:", user_options)
camera_cfg_file = user_options['camera_config']
if camera_cfg_file != "":
camera_cfg_file += ".cfg"
cam_config = ConfigParser.ConfigParser()
cam_cfg_read = cam_config.read(camera_cfg_file)
if cam_cfg_read != []:
cam_options = ConfigSectionMap(cam_config, "Resolution")
try:
cam_width = int(cam_options['width'])
cam_height = int(cam_options['height'])
except ValueError:
print("Height or Width setting invalid!")
else:
print(
"Invalid Camera_Config specified in VidMan.cfg, section [USER] - using default frame size 1920 x 1080"
)
else:
print(
"No Camera_Config specified in VidMan.cfg, section [USER] - using default frame size 1920 x 1080"
)
crop_height_str = user_options['crop_height']
try:
crop_height = int(crop_height_str)
except ValueError:
crop_height = cam_height
crop_width_str = user_options['crop_width']
try:
crop_width = int(crop_width_str)
except ValueError:
crop_width = cam_width
crop_width = min(crop_width, cam_width)
crop_height = min(crop_height, cam_height)
# Set the width and height to maximum
# Logitech C920:
# cap.set(cv2.cv.CV_CAP_PROP_FRAME_WIDTH, cam_width)
# cap.set(cv2.cv.CV_CAP_PROP_FRAME_HEIGHT, cam_height)
# Logitech Quickcam Pro 9000:
# cap.set(cv2.cv.CV_CAP_PROP_FRAME_WIDTH, 1600)
# cap.set(cv2.cv.CV_CAP_PROP_FRAME_HEIGHT, 1200)
# Ipevo Ziggi HD:
cap.set(cv2.cv.CV_CAP_PROP_FRAME_WIDTH, 1920)
cap.set(cv2.cv.CV_CAP_PROP_FRAME_HEIGHT, 1080)
img = np.zeros((cam_height, cam_width, 3),
np.uint8) # initialise img as blank
cv2.putText(img, "No camera", (50, 100), cv2.FONT_HERSHEY_SIMPLEX, 5.0,
COLOUR_GREEN, 2)
while True:
if not freeze: # "live" mode - grab a new image on each pass round this loop
ret, img = cap.read()
# print ret
working_img = img # save creating a new object - directly modify the newly-grabbed frame
else: # freeze-frame
if take_new_frame:
take_new_frame = False
for i in range(
0, 5
): # (need to flush some buffered frames from the camera)
ret, img = cap.read()
working_img = copy.copy(
img
) # keep the grabbed frame unmodified, create a copy to work on
# crop the captured frame, if required:
if (crop_width < cam_width) or (crop_height < cam_height):
working_img = working_img[0.5 * (cam_height - crop_height):0.5 *
(cam_height + crop_height),
0.5 * (cam_width - crop_width):0.5 *
(cam_width + crop_width)]
if rotate180:
working_img = cv2.flip(
working_img, -1
) # compensate for webcam being mounted effectively upside-down
if magnification > 1.0:
height, width = working_img.shape[:2]
cropped_img = working_img[0.5 * height *
(1 - 1 / magnification):0.5 * height *
(1 + 1 / magnification), 0.5 * width *
(1 - 1 / magnification):0.5 * width *
(1 + 1 / magnification)]
#working_img = cv2.resize(cropped_img, None, fx=magnification, fy=magnification, \
# interpolation = cv2.INTER_CUBIC)
working_img = cv2.resize(cropped_img, None, fx=magnification, fy=magnification, \
interpolation=cv2.INTER_LINEAR)
#OR
#height, width = cropped_img.shape[:2]
#res = cv2.resize(cropped_img, (magnification*width, magnification*height), interpolation = cv2.INTER_CUBIC)
if brightness != 1.0 or contrast != 1.0 or sharpness != 1.0:
pil_image = Image.fromarray(
working_img) # convert OpenCV image to PIL format
if brightness != 1.0:
enh = ImageEnhance.Brightness(pil_image)
pil_image = enh.enhance(brightness)
if contrast != 1.0:
enh = ImageEnhance.Contrast(pil_image)
pil_image = enh.enhance(contrast)
if sharpness != 1.0:
enh = ImageEnhance.Sharpness(pil_image)
pil_image = enh.enhance(sharpness)
working_img = np.array(
pil_image) # convert PIL-format image back to OpenCV format
if autocontrast:
grayscale_img = cv2.cvtColor(working_img, cv2.COLOR_BGR2GRAY)
working_img = cv2.adaptiveThreshold(grayscale_img, 255, cv2.ADAPTIVE_THRESH_GAUSSIAN_C,\
cv2.THRESH_BINARY, 29, 2)
if black_on_yellow:
working_img = cv2.cvtColor(
working_img,
cv2.COLOR_GRAY2RGB) # convert back to RGB image
if black_on_yellow:
# note that [:,:,0] is blue, [:,:,1] is green, [:,:,2] is red
working_img[:, :, 0] = 0
if inverted:
working_img = 255 - working_img
if autofocus:
focusval = UVC_functions.get_focus()
if show_params:
if autocontrast:
cv2.putText(working_img, "Autocontrast on", (20, 30),
cv2.FONT_HERSHEY_SIMPLEX, 1.0, COLOUR_WHITE, 2)
cv2.putText(working_img, "Autocontrast on", (20, 70),
cv2.FONT_HERSHEY_SIMPLEX, 1.0, COLOUR_BLACK, 2)
else:
cv2.putText(working_img, "Brightness: " + str(brightness),
(20, 30), cv2.FONT_HERSHEY_SIMPLEX, 1.0,
COLOUR_YELLOW, 2)
cv2.putText(working_img, "Contrast: " + str(contrast),
(20, 70), cv2.FONT_HERSHEY_SIMPLEX, 1.0,
COLOUR_YELLOW, 2)
cv2.putText(working_img, "Sharpness: " + str(sharpness),
(20, 110), cv2.FONT_HERSHEY_SIMPLEX, 1.0,
COLOUR_YELLOW, 2)
cv2.putText(working_img, "Focus: " + str(focusval), (20, 150),
cv2.FONT_HERSHEY_SIMPLEX, 1.0, COLOUR_YELLOW, 2)
cv2.imshow("Video Magnifier", working_img)
if not freeze:
key_wait_time = 100 # image live-update mode - wait 100ms
else:
key_wait_time = 0 # i.e. if frame is frozen, wait indefinitely for a key to be pressed
key = cv2.waitKey(
key_wait_time) % 256 # LSB equates to ASCII code for most keys
# print(key)
if key == 27: # Esc - quit the program
break
elif key == ord('1'): # 1 - normal colours
inverted = False
elif key == ord('2'): # 2 - colours inverted
inverted = True
elif key == ord('3'): # 3 - auto-contrast off
autocontrast = False
elif key == ord('4'): # 4 - auto-contrast on
autocontrast = True
elif key == ord('5'): # 5 - black-on-yellow mode off
black_on_yellow = False
elif key == ord('6'): # 6 - black-on-yellow mode on
black_on_yellow = True
elif key == 176: # num-pad 0 - decrease sharpness
sharpness -= 0.1
if sharpness < 0.0:
sharpness = 0.0
elif key == 174: # num-pad . - increase sharpness
sharpness += 0.1
if sharpness > 10.0:
sharpness = 10.0
elif key == 177: # num-pad 1 - 100% zoom
magnification = 1.0
elif key == 178: # num-pad 2 - 200% zoom
magnification = 2.0
elif key == 179: # num-pad 3 - 300% zoom
magnification = 3.0
elif key == 227: # left-Ctrl - unfreeze frame
freeze = False
elif key == 82: # up-arrow - increase brightness
brightness += 0.1
if brightness > 10.0:
brightness = 10.0
elif key == 84: # down-arrow - decrease brightness
brightness -= 0.1
if brightness < 0.0:
brightness = 0.0
elif key == 81: # left-arrow - decrease contrast
contrast -= 0.1
if contrast < 0.0:
contrast = 0.0
elif key == 83: # right-arrow - increase contrast
contrast += 0.1
if contrast > 10.0:
contrast = 10.0
elif key == 80: # Home key - reset to default colours
brightness = 1.0
contrast = 1.0
sharpness = 1.0
inverted = False
autocontrast = False
black_on_yellow = False
elif key == 32: # space-bar - freeze frame, or grab a new frozen frame
if freeze:
take_new_frame = True
freeze = True
elif key == 190: # F1 - don't show parameter text overlay
show_params = False
elif key == 191: # F2 - show parameters overlaid on image
show_params = True
elif key == 141: # Num Enter - autofocus on
UVC_functions.set_autofocus(True)
autofocus = True
elif key == 171: # Num + - autofocus off, increase focus value
if autofocus:
UVC_functions.set_autofocus(False)
autofocus = False
focusval += 5
UVC_functions.set_focus(focusval)
elif key == 173: # Num - - autofocus off, decrease focus value
if autofocus:
UVC_functions.set_autofocus(False)
autofocus = False
focusval -= 5
UVC_functions.set_focus(focusval)
def filterr(img):
"""takes a picture and puts a fliter on it"""
try:
print(
"""1.BLUR\t\t\t16.NEGATIVE LESS\t30.GINGHAM\n2.DETAIL\t\t17.NEGATIVE MORE\t31.ARTIC\n3.CONTOUR\t\t18.EQUALIZE\t\t32.DIBUJO\n4.EDGE_ENHANCE\t\t19.MORE BRIGHT\t\t33.CHARMES
5.EDGE_ENHANCE_MORE\t20.AUTOCONTRAST\n6.EMBOSS\t\t21.MORE CONTRAST\n7.FIND_EDGES\t\testos tardan un poco mas\n8.SMOOTH\t\t22.BLACK AND WHITE LESS
9.SMOOTH_MORE\t\t23.THRESHOLDS\n10.SHARPEN\t\t24.THRESHOLDS LESS\n11.SHARPEN MORE\t\t25.AVERAGE\n12.GRAYSCALE\t\t26.AVERAGE ALL NEGHBORS\n13.POSTERIZE\t\t27.OLD PICTURE
14.SOLARIZE\t\t28.ORANGE\n15.NEGATIVE\t\t29.VALENCIA""")
i = input("Seleccione una opcion:")
if i == "1":
filpic = img.filter(ImageFilter.BLUR)
filpic.show()
elif i == "2":
filpic = img.filter(ImageFilter.DETAIL)
filpic.show()
elif i == "3":
filpic = img.filter(ImageFilter.CONTOUR)
filpic.show()
elif i == "4":
filpic = img.filter(ImageFilter.EDGE_ENHANCE)
filpic.show()
elif i == "5":
filpic = img.filter(ImageFilter.EDGE_ENHANCE_MORE)
filpic.show()
elif i == "6":
filpic = img.filter(ImageFilter.EMBOSS)
filpic.show()
elif i == "7":
filpic = img.filter(ImageFilter.FIND_EDGES)
filpic.show()
elif i == "8":
filpic = img.filter(ImageFilter.SMOOTH)
filpic.show()
elif i == "9":
filpic = img.filter(ImageFilter.SMOOTH_MORE)
filpic.show()
elif i == "10":
filpic = img.filter(ImageFilter.SHARPEN)
filpic.show()
elif i == "11":
filpic = ImageEnhance.Sharpness(img).enhance(9)
filpic.show()
elif i == "12":
filpic = ImageOps.grayscale(img)
filpic.show()
elif i == "13":
filpic = ImageOps.posterize(img, 1)
filpic.show()
elif i == "14":
filpic = ImageOps.solarize(img)
filpic.show()
elif i == "15":
filpic = ImageOps.invert(img)
filpic.show()
elif i == "16":
filpic = ImageEnhance.Contrast(img).enhance(-2)
filpic.show()
elif i == "17":
filpic = ImageEnhance.Contrast(img).enhance(-7)
filpic.show()
elif i == "18":
filpic = ImageOps.equalize(img)
filpic.show()
elif i == "19":
filpic = ImageEnhance.Brightness(img).enhance(2)
filpic.show()
elif i == "20":
filpic = ImageOps.autocontrast(img, cutoff=2, ignore=None)
filpic.show()
elif i == "21":
filpic = ImageEnhance.Contrast(img).enhance(2)
filpic.show()
elif i == "22":
black_and_white_less(img)
elif i == "23":
thresholds(img)
elif i == "24":
thresholdsless(img)
elif i == "25":
average(img)
elif i == "26":
average_allneighbors(img)
elif i == "27":
sepia(img)
elif i == "28":
sepiaDF(img)
elif i == "29":
sepiaDFlessD(img)
elif i == "30":
sepiaDFless(img)
elif i == "31":
print("Se puede mejorar añadiendo Contraste")
juno(img)
elif i == "32":
juno2(img)
elif i == "33":
notgray(img)
else:
print("Esa no vale")
if int(i) <= 21 and int(i) != 0:
yn = input("Quieres guardar(s/n):")
saveimg(yn, filpic)
except:
print("No se pudo realizar la operacion")
def align_dataset_mtcnn_url():
rs_username = request.json['txtusername']
directory = rs_username.replace(" ", "")
parent_dir = os.path.expanduser("Dataset/FaceData/raw")
path = os.path.join(parent_dir, directory)
try:
os.mkdir(path)
print("Directory '% s' created" % directory)
except OSError as err:
message = "OS error: {0}".format(err)
return make_response(
jsonify(
{"message": message}
),
400,
)
url_img_list = request.json['url_list']
if len(url_img_list) != 20:
return make_response(
jsonify(
{"message": "Need 20 images for tranning"}
),
400,
)
UPLOAD_FOLDER = os.path.expanduser("Dataset/FaceData/raw") + '/' + directory
app.config['UPLOAD_FOLDER'] = UPLOAD_FOLDER
for i, url_img in enumerate(url_img_list):
filename = 'image' + str(i) + '.jpg'
urllib.request.urlretrieve(url_img, os.path.join(app.config['UPLOAD_FOLDER'], filename))
flash('File successfulliy uploaded ' + filename + ' to the database!')
os.system("python src/align_dataset_mtcnn.py Dataset/FaceData/raw"
+ " Dataset/FaceData/processed --image_size 160 "
+ "--margin 32 --random_order --gpu_memory_fraction 0.25")
output_dir = os.path.expanduser("Dataset/FaceData/processed/")
if not os.path.exists(output_dir):
os.makedirs(output_dir)
nrof_images_total = 0
output_class_dir = os.path.join(output_dir, directory)
if not os.path.exists(output_class_dir):
os.makedirs(output_class_dir)
for image_path in src.facenet.get_imgDirs("Dataset/FaceData/processed/" + directory):
nrof_images_total += 1
filename = os.path.splitext(os.path.split(image_path)[1])[0] + str(nrof_images_total)
output_filename = os.path.join(output_class_dir, filename + '.png')
print(image_path)
image_obj = Image.open(image_path);
rotated_image2 = image_obj.convert('L')
rotated_image2.save(output_filename)
image_flip = image_obj.transpose(Image.FLIP_LEFT_RIGHT)
output_filename = os.path.join(output_class_dir, filename + 'a.png')
image_flip.save(output_filename);
output_filename = os.path.join(output_class_dir, filename + 'b.png')
sharpness = ImageEnhance.Sharpness(image_obj);
sharpness.enhance(1.5).save(output_filename);
output_filename = os.path.join(output_class_dir, filename + 'c.png')
color = ImageEnhance.Color(image_obj)
color.enhance(1.5).save(output_filename)
output_filename = os.path.join(output_class_dir, filename + 'd.png')
brightness = ImageEnhance.Brightness(image_obj)
brightness.enhance(1.5).save(output_filename);
return make_response(
jsonify(
{"message": "Success"}
),
200,
)
def __init__(self, p1, operation1, magnitude_idx1, p2, operation2, magnitude_idx2, fillcolor=(128, 128, 128)):
ranges = {
"shearX": np.linspace(0, 0.3, 10),
"shearY": np.linspace(0, 0.3, 10),
"translateX": np.linspace(0, 150 / 331, 10),
"translateY": np.linspace(0, 150 / 331, 10),
"rotate": np.linspace(0, 30, 10),
"color": np.linspace(0.0, 0.9, 10),
"posterize": np.round(np.linspace(8, 4, 10), 0).astype(np.int),
"solarize": np.linspace(256, 0, 10),
"contrast": np.linspace(0.0, 0.9, 10),
"sharpness": np.linspace(0.0, 0.9, 10),
"brightness": np.linspace(0.0, 0.9, 10),
"autocontrast": [0] * 10,
"equalize": [0] * 10,
"invert": [0] * 10
}
# from https://stackoverflow.com/questions/5252170/specify-image-filling-color-when-rotating-in-python-with-pil-and-setting-expand
def rotate_with_fill(img, magnitude):
rot = img.convert("RGBA").rotate(magnitude)
return Image.composite(rot, Image.new("RGBA", rot.size, (128,) * 4), rot).convert(img.mode)
func = {
"shearX": lambda img, magnitude: img.transform(
img.size, Image.AFFINE, (1, magnitude * random.choice([-1, 1]), 0, 0, 1, 0),
Image.BICUBIC, fillcolor=fillcolor),
"shearY": lambda img, magnitude: img.transform(
img.size, Image.AFFINE, (1, 0, 0, magnitude * random.choice([-1, 1]), 1, 0),
Image.BICUBIC, fillcolor=fillcolor),
"translateX": lambda img, magnitude: img.transform(
img.size, Image.AFFINE, (1, 0, magnitude * img.size[0] * random.choice([-1, 1]), 0, 1, 0),
fillcolor=fillcolor),
"translateY": lambda img, magnitude: img.transform(
img.size, Image.AFFINE, (1, 0, 0, 0, 1, magnitude * img.size[1] * random.choice([-1, 1])),
fillcolor=fillcolor),
"rotate": lambda img, magnitude: rotate_with_fill(img, magnitude),
# "rotate": lambda img, magnitude: img.rotate(magnitude * random.choice([-1, 1])),
"color": lambda img, magnitude: ImageEnhance.Color(img).enhance(1 + magnitude * random.choice([-1, 1])),
"posterize": lambda img, magnitude: ImageOps.posterize(img, magnitude),
"solarize": lambda img, magnitude: ImageOps.solarize(img, magnitude),
"contrast": lambda img, magnitude: ImageEnhance.Contrast(img).enhance(
1 + magnitude * random.choice([-1, 1])),
"sharpness": lambda img, magnitude: ImageEnhance.Sharpness(img).enhance(
1 + magnitude * random.choice([-1, 1])),
"brightness": lambda img, magnitude: ImageEnhance.Brightness(img).enhance(
1 + magnitude * random.choice([-1, 1])),
"autocontrast": lambda img, magnitude: ImageOps.autocontrast(img),
"equalize": lambda img, magnitude: ImageOps.equalize(img),
"invert": lambda img, magnitude: ImageOps.invert(img)
}
# self.name = "{}_{:.2f}_and_{}_{:.2f}".format(
# operation1, ranges[operation1][magnitude_idx1],
# operation2, ranges[operation2][magnitude_idx2])
self.p1 = p1
self.operation1 = func[operation1]
self.magnitude1 = ranges[operation1][magnitude_idx1]
self.p2 = p2
self.operation2 = func[operation2]
self.magnitude2 = ranges[operation2][magnitude_idx2]
def __init__(self, Numbers=None, max_Magnitude=None, **kwargs):
'''
Custom image data generator.
Behaves like ImageDataGenerator, but allows color augmentation.
'''
super().__init__(preprocessing_function=self.__call__, **kwargs)
self.transforms = ['autocontrast', 'equalize', 'rotate', 'solarize', 'color', 'posterize',
'contrast', 'brightness', 'sharpness', 'shearX', 'shearY', 'translateX', 'translateY']
if Numbers is None:
self.Numbers = len(self.transforms) // 2
else:
self.Numbers = Numbers
if max_Magnitude is None:
self.max_Magnitude = 10
else:
self.max_Magnitude = max_Magnitude
fillcolor = 128
self.ranges = {
# these Magnitude range , you must test it yourself , see what will happen after these operation ,
# it is no need to obey the value in autoaugment.py
"shearX": np.linspace(0, 0.3, 10),
"shearY": np.linspace(0, 0.3, 10),
"translateX": np.linspace(0, 0.2, 10),
"translateY": np.linspace(0, 0.2, 10),
"rotate": np.linspace(0, 360, 10),
"color": np.linspace(0.0, 0.9, 10),
"posterize": np.round(np.linspace(8, 4, 10), 0).astype(np.int),
"solarize": np.linspace(256, 231, 10),
"contrast": np.linspace(0.0, 0.5, 10),
"sharpness": np.linspace(0.0, 0.9, 10),
"brightness": np.linspace(0.0, 0.3, 10),
"autocontrast": [0] * 10,
"equalize": [0] * 10,
"invert": [0] * 10
}
self.func = {
"shearX": lambda img, magnitude: img.transform(
img.size, Image.AFFINE, (1, magnitude * random.choice([-1, 1]), 0, 0, 1, 0),
Image.BICUBIC, fill=fillcolor),
"shearY": lambda img, magnitude: img.transform(
img.size, Image.AFFINE, (1, 0, 0, magnitude * random.choice([-1, 1]), 1, 0),
Image.BICUBIC, fill=fillcolor),
"translateX": lambda img, magnitude: img.transform(
img.size, Image.AFFINE, (1, 0, magnitude * img.size[0] * random.choice([-1, 1]), 0, 1, 0),
fill=fillcolor),
"translateY": lambda img, magnitude: img.transform(
img.size, Image.AFFINE, (1, 0, 0, 0, 1, magnitude * img.size[1] * random.choice([-1, 1])),
fill=fillcolor),
"rotate": lambda img, magnitude: self.rotate_with_fill(img, magnitude),
# "rotate": lambda img, magnitude: img.rotate(magnitude * random.choice([-1, 1])),
"color": lambda img, magnitude: ImageEnhance.Color(img).enhance(1 + magnitude * random.choice([-1, 1])),
"posterize": lambda img, magnitude: ImageOps.posterize(img, magnitude),
"solarize": lambda img, magnitude: ImageOps.solarize(img, magnitude),
"contrast": lambda img, magnitude: ImageEnhance.Contrast(img).enhance(
1 + magnitude * random.choice([-1, 1])),
"sharpness": lambda img, magnitude: ImageEnhance.Sharpness(img).enhance(
1 + magnitude * random.choice([-1, 1])),
"brightness": lambda img, magnitude: ImageEnhance.Brightness(img).enhance(
1 + magnitude * random.choice([-1, 1])),
"autocontrast": lambda img, magnitude: ImageOps.autocontrast(img),
"equalize": lambda img, magnitude: img,
"invert": lambda img, magnitude: ImageOps.invert(img)
}
def test_crash():
# crashes on small images
im = Image.new("RGB", (1, 1))
ImageEnhance.Sharpness(im).enhance(0.5)
for j in range(regen):
z = np.random.normal(loc=0.0, scale=1.0, size=(batch_size, z_dim)).astype(np.float32)
gen = sess.run(net_g.out, feed_dict={t_real_caption: all_cap, t_z: z})
score = sess.run(disc.out, feed_dict={fake_image: gen,t_real_caption: all_cap}).reshape([-1])
for k in range(batch_size):
if score[k]<best_score[k]:
best_imgs[k,:,:,:] = gen[k,:,:,:]
best_score[k] = score[k]
for j in range(batch_size):
image_path = 'inference/inference_'+test_sentence['ID'].values[i*batch_size+j]+'.png'
images = best_imgs[j]*0.5+0.5
#########################################
img = Image.fromarray(np.uint8(images*255))
enhancer = ImageEnhance.Sharpness(img)
img = enhancer.enhance(3.0)
#don't use color enhance
# enhancer = ImageEnhance.Color(img)
# images = np.array(enhancer.enhance(1.3))
#########################################
scipy.misc.imsave(image_path,(img))
################################### deal with last batch
all_cap = []
for i in range(n_caption_test-batch_size,n_caption_test):
test_cap = test_sentence['Captions'].values[i]
def sharpness(img, factor, **__):
return ImageEnhance.Sharpness(img).enhance(factor)
img_denoised = np.logical_not(sgm).astype(np.uint8) * 255
# Save denoised image
cv2.imwrite(Working_Path + '\\' + 'Denoised_' + index, img_denoised)
# Load image, grayscale, Gaussian blur, adaptive threshold
img = cv2.imread(Working_Path + '\\' + 'Denoised_' + index)
ret, thresh = cv2.threshold(img, 55, 255, cv2.THRESH_BINARY)
opening = cv2.morphologyEx(
thresh, cv2.MORPH_OPEN,
cv2.getStructuringElement(cv2.MORPH_RECT, (2, 2)))
cv2.imwrite(Procees_Path + '\\' + 'Enhancexd_' + index, opening)
#Enhance The Image
image = Image.open(Procees_Path + '\\' + 'Enhancexd_' + index)
enh_sha = ImageEnhance.Sharpness(image)
sharpness = 100
image_sharped = enh_sha.enhance(sharpness)
image4 = np.array(image_sharped)
cv2.imwrite(Final_Input + '\\' + 'Final_' + index, image4)
# ********************************************************************************************************************************
#OCR Part
# ********************************************************************************************************************************
im = Image.open(Final_Input + '\\' + 'Final_' + index) # the second one
im = im.filter(ImageFilter.ModeFilter()) #Filter The Image
enhancer = ImageEnhance.Sharpness(im)
im = enhancer.enhance(4)
im = im.convert('1')
im.save('temp2.jpg')
def sharpness(pil_img, level):
level = float_parameter(sample_level(level), 1.8) + 0.1
return ImageEnhance.Sharpness(pil_img).enhance(level)
def __init__(self,
dataset,
transformations=None,
n_classes=8000,
n_trans=100,
max_elms=10,
p=0.5):
"""ExemplarNet dataset.
Args:
dataset (torch.utils.data.Dataset): The dataset to train on.
transformations (list, optional): Type of elementar transformations to use.
n_classes (int, optional): Number of classes, i.e. the subset size of the dataset. Defaults to 8000.
n_trans (int, optional): Number of combined transformations. Defaults to 100.
max_elms (int, optional): Number of elementar transformations per combined transformation. Defaults to 10.
p (float, optional): Prob. of an elmentar transformation to be part of a combined transformation. Defaults to 0.5.
"""
pool = [
transforms.RandomRotation( # Rotation
30,
resample=False,
expand=False,
center=None,
fill=None),
transforms.RandomAffine( # Shearing
0,
translate=None,
scale=None,
shear=30,
resample=False,
fillcolor=0),
transforms.RandomAffine( # Translate
0,
translate=(0.3, 0.3),
scale=None,
shear=None,
resample=False,
fillcolor=0),
transforms.Lambda(lambda x: imo.autocontrast(x)), # Autocontrast
transforms.Lambda(lambda x: imo.invert(x)), # Invert
transforms.Lambda(lambda x: imo.equalize(x)), # Equalize
transforms.Lambda(lambda x: imo.solarize(x)), # Solarize
transforms.Lambda(lambda x: imo.posterize(
x, bits=int(np.random.randint(4, 8) + 1))), # Posterize
transforms.Lambda(
lambda x: ime.Color(x).enhance(np.random.uniform())), # Color
transforms.Lambda(lambda x: ime.Brightness(x).enhance(
np.random.uniform())), # Brightness
transforms.Lambda(lambda x: ime.Contrast(x).enhance(
np.random.uniform())), # Contrast
transforms.Lambda(lambda x: ime.Sharpness(x).enhance(
np.random.uniform())), # Sharpness
transforms.Compose( # Set black
[
transforms.ToTensor(),
transforms.RandomErasing(1.0),
transforms.ToPILImage()
]),
transforms.Lambda(
lambda x: transforms.functional.to_grayscale( # Grayscale
x, num_output_channels=3)),
transforms.Lambda(lambda x: elastic_transform(x, sigma=10))
]
# Processes full images and apply random cropping instead of gradient based sampling.
indices = torch.randint(len(dataset), (n_classes, )).long()
self.dataset = Subset(dataset, indices)
self.p = p
self.n_trans = n_trans
elm_transformations = transformations if transformations is not None else pool
self.transformations = []
for _ in range(self.n_trans):
transformation = []
for t in range(max_elms):
if random.random() < self.p:
transformation.append(
transforms.RandomChoice(elm_transformations))
self.transformations.append(transforms.Compose(transformation))
def main():
#default device
device = '/device:CPU:0'
parser = argparse.ArgumentParser()
parser.add_argument('--model-name',
help='Model name to use for classification',
type=str)
parser.add_argument('--captcha-dir',
help='Where to read the captchas to break',
type=str)
parser.add_argument('--output',
help='File where the classifications should be saved',
type=str)
parser.add_argument('--symbols',
help='File with the symbols to use in captchas',
type=str)
parser.add_argument('--gpu', help='used to run in gpu', type=str)
args = parser.parse_args()
if args.model_name is None:
print("Please specify the CNN model to use")
exit(1)
if args.captcha_dir is None:
print("Please specify the directory with captchas to break")
exit(1)
if args.output is None:
print("Please specify the path to the output file")
exit(1)
if args.symbols is None:
print("Please specify the captcha symbols file")
exit(1)
if args.gpu == 'gpu':
physical_devices = tf.config.experimental.list_physical_devices('GPU')
assert len(physical_devices) > 0, "No GPU available!"
tf.config.experimental.set_memory_growth(physical_devices[0], True)
device = '/device:GPU:0'
symbols_file = open(args.symbols, 'r')
captcha_symbols = symbols_file.readline().strip()
symbols_file.close()
print("Classifying captchas with symbol set {" + captcha_symbols + "}")
with tf.device(device):
with open(args.output, 'w') as output_file:
json_file = open(args.model_name + '.json', 'r')
loaded_model_json = json_file.read()
json_file.close()
model = keras.models.model_from_json(loaded_model_json)
model.load_weights(args.model_name + '.h5')
model.compile(loss='categorical_crossentropy',
optimizer=keras.optimizers.Adam(1e-3, amsgrad=True),
metrics=['accuracy'])
for x in os.listdir(args.captcha_dir):
# load image and preprocess it
image = Image.open(os.path.join(args.captcha_dir, x))
image = ImageOps.autocontrast(image, cutoff=10, ignore=None)
image = ImageEnhance.Sharpness(image)
image = image.enhance(10.0)
image = ImageOps.grayscale(image)
image = numpy.array(image)
image = cv2.threshold(image, 0, 255,
cv2.THRESH_BINARY + cv2.THRESH_OTSU)[1]
image = numpy.array(image) / 255.0
image = numpy.expand_dims(image, axis=2)
(c, h, w) = image.shape
image = image.reshape([-1, c, h, w])
prediction = model.predict(image)
output_file.write(x + "," +
decode(captcha_symbols, prediction) + "\n")
print('Classified ' + x)
print("Classifying captchas completed!")
def sharpness(img, magnitude):
magnitudes = np.linspace(0.1, 1.9, 11)
img = ImageEnhance.Sharpness(img).enhance(random.uniform(magnitudes[magnitude], magnitudes[magnitude+1]))
return img