def classify_image(images, mask_list, k_size, save, display):
"""
Classify pixels of a single image
"""
if len(images) > 1:
raise ValueError('Only one image can be classified at once')
logging.info('Calculating, normalizing feature vectors for image')
image = images[0] # First and only member
vectors = calculate_features(image.image, image.fov_mask, mask_list,
k_size)
logging.info('Classifying image pixels')
probabilities, prediction = svm.classify(vectors)
svm.assess(image.truth, prediction)
svm.plot_roc(image.truth, probabilities)
if save:
image_utils.save_image(prediction, 'prediction.png')
logging.info('Saved classified image')
if display:
image_utils.display_image(prediction)
logging.info('Displaying classified image')
def split_squares_average(img: np.ndarray, n_squares: int, marker_size: int):
"""
Split the image in squares
:param img: the original image
:param n_squares: the number of squares
:param marker_size: the width of the line in pixels
"""
result_img = img.copy()
square_size = img.shape[0] // n_squares
for nx in range(n_squares - 1):
for ny in range(n_squares - 1):
current_cell = result_img[square_size * nx:square_size * (nx + 1),
square_size * ny:square_size *
(ny + 1), :]
for ch in range(3):
result_img[square_size*nx:square_size*(nx+1), square_size*ny:square_size*(ny+1), ch] = \
np.mean(current_cell[:, :, ch])
for n in range(n_squares):
result_img[square_size * n:square_size * n + 5, :, :] = 255
result_img[:, square_size * n:square_size * n + 5, :] = 255
display_image(result_img, title="test", cmap=None)
def split_squares_random_direction_sorting(img: np.ndarray, n_squares: int,
marker_size: int):
"""
Split the image in squares
:param img: the original image
:param n_squares: the number of squares
:param marker_size: the width of the line in pixels
"""
result_img = img.copy()
square_size = img.shape[0] // n_squares
for nx in range(n_squares - 1):
for ny in range(n_squares - 1):
current_cell_pixels = result_img[square_size * nx:square_size *
(nx + 1), square_size *
ny:square_size * (ny + 1), :]
result_img[square_size*nx:square_size*(nx+1), square_size*ny:square_size*(ny+1), :] = \
pixel_sorting(current_cell_pixels, direction="UP")
for n in range(n_squares):
result_img[square_size * n:square_size * n + 5, :, :] = 255
result_img[:, square_size * n:square_size * n + 5, :] = 255
display_image(result_img, title="rotation test", cmap=None)
multiple_folders = []
for folder in count_by_folders:
if count_by_folders[folder] == 1:
singleton_folders.append(folder)
elif count_by_folders[folder] > 1:
multiple_folders.append(folder)
else:
print("EMPTY FOLDER: ", folder)
def get_batch():
number_of_pairs = ((batch_size + 1) // 2) // 2
singles = sample(singleton_folders, k=batch_size - number_of_pairs * 2)
multiples = sample(multiple_folders, k=number_of_pairs)
batch = []
for m in multiples:
ms = sample(image_paths[m], k=2)
# print(ms)
for filename in ms:
batch.append(image_utils.load_image(os.path.join(m, filename), training_params.shape))
for s in singles:
batch.append(image_utils.load_image(os.path.join(s, image_paths[s][0]), training_params.shape))
return np.array(batch)
if __name__ == '__main__':
batch = get_batch()
print(batch.shape)
image_utils.display_image(image_utils.unpreprocess_image(batch[0]))
image_utils.display_image(image_utils.unpreprocess_image(batch[1]))
temp_mi_loss = 20 * max(
-1 * np.average(np.square(mi[i] - mi[j])) + 2., 0)
# print(i, ",", j, '-NOT A PAIR', temp_mi_loss, np.sum(np.square(mi[i], np.zeros_like(mi[i])))**0.5)
# print(np.sum(np.square(mi[i] - mi[j]))**0.5)
running_mi_loss += temp_mi_loss
print(
"MI LOSS: ", running_mi_loss / ((params.batch_size *
(params.batch_size - 1)) / 2))
o, lgx = sess.run([output, loggamma_x],
feed_dict={
X: next_batch,
training: True
})
print(o.shape)
for j in range(params.batch_size):
image_utils.display_image(image_utils.unpreprocess_image(o[j]))
image_utils.display_image(
image_utils.unpreprocess_image(next_batch[j]))
print(
"IMAGE ", j, ": ",
np.average(
np.power(10 *
(o[j] - next_batch[j]) / np.exp(lgx), 2) /
2.0 + lgx))
diff_img = (np.power(o[j] - next_batch[j], 2) - 2.0) / 2.0
image_utils.display_image(
image_utils.unpreprocess_image(diff_img))
if j > 1:
break
print(
"TOTAL LOSS: ",
def get_image_samples(query):
image_type = "ActiOn"
web_query = (query + ' face').split()
web_query = '+'.join(web_query)
url = "https://www.google.co.in/search?q=" + web_query + "&source=lnms&tbm=isch"
print(url)
#add the directory for your image here
DIR = "Pictures"
header = {
'User-Agent':
"Mozilla/5.0 (Windows NT 6.1; WOW64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/43.0.2357.134 Safari/537.36"
}
soup = get_soup(url, header)
ActualImages = [
] # contains the link for Large original images, type of image
for a in soup.find_all("div", {"class": "rg_meta"}):
link, Type = json.loads(a.text)["ou"], json.loads(a.text)["ity"]
if len(ActualImages) < 10:
ActualImages.append((link, Type))
print("there are total", len(ActualImages), "images")
if not os.path.exists(DIR):
os.mkdir(DIR)
DIR = os.path.join(DIR, web_query.split()[0])
if not os.path.exists(DIR):
os.mkdir(DIR)
###print images
for i, (img, Type) in enumerate(ActualImages):
try:
req = urllib2.Request(img, headers={'User-Agent': header})
raw_img = urllib2.urlopen(req.full_url).read()
cntr = len([i for i in os.listdir(DIR) if image_type in i]) + 1
print(cntr)
if len(Type) == 0:
filename = os.path.join(DIR,
image_type + "_" + str(cntr) + ".jpg")
f = open(filename, 'wb')
else:
filename = os.path.join(
DIR, image_type + "_" + str(cntr) + "." + Type)
f = open(filename, 'wb')
f.write(raw_img)
f.close()
target_path = os.path.join(image_dir, "_".join(query.split(' ')),
'supplement.jpg')
print("TARGET_PATH: ", target_path)
display_image(unpreprocess_image(load_image(filename)))
for root, dir, files in os.walk(
os.path.join(image_dir, "_".join(query.split(' ')))):
for file in files:
display_image(
unpreprocess_image(load_image(os.path.join(root,
file))))
ans = input("KEEP THIS IMAGE?: ")
if ans == 'y':
os.rename(filename, target_path)
break
except Exception as e:
print("could not load : " + img)
print(e)
import image_utils
import training_utils
import numpy as np
image_color = image_utils.load_image('test samples/test_samples(0,25).png')
img = image_utils.image_bin(image_utils.image_gray(image_color))
img_bin = image_utils.erode(image_utils.dilate(img))
alphabet = []
selected_regions, numbers, alphabet = image_utils.select_roi(
image_color.copy(), img, alphabet)
image_utils.display_image(selected_regions)
print(alphabet)
inputs = image_utils.prepare_for_ann(numbers)
outputs = training_utils.convert_output(alphabet)
ann = training_utils.load_modell()
#print(outputs)
result = ann.predict(np.array(inputs, np.float32))
final_alphabet = training_utils.convert_output([0, 1, 2, 3, 4, 5, 6, 7, 8, 9])
print(result)
print(training_utils.display_result(result, final_alphabet, alphabet))
elif direction == "DOWN":
pixel_list = img[e[0]:, e[1], :]
sorted_pixels = sorted(pixel_list, key=lambda x: x[2])
result[e[0]:, e[1], :] = sorted_pixels[::-1]
elif direction == "LEFT":
# TODO "Fix value error exception"
pixel_list = img[e[0], :e[1], :]
sorted_pixels = sorted(pixel_list, key=lambda x: x[2])
result[e[0], :e[1], :] = sorted_pixels[::-1]
elif direction == "RIGHT":
pixel_list = img[e[0], e[1]:, :]
sorted_pixels = sorted(pixel_list, key=lambda x: x[2])
result[e[0], e[1]:, :] = sorted_pixels[::-1]
return result
if __name__ == '__main__':
for i in range(2, 3):
fname = str(i) + ".jpg"
file_name, image = load_image(fname, cv2_read_param=1)
display_image(image, title="Original image", cmap=None)
edgemap = cv2.Canny(image, 100, 500)
# enhanced_borders = enhance_borders(image, edgemap, marker_size=2)
# display_image(enhanced_borders, title="Canny", cmap=None)
# result_img = pixel_propagation_along_edges(image, edgemap, "RIGHT")
result_img = pixel_sorting_along_edges(image, edgemap, "DOWN")
# result_img = pixel_sorting_along_edges(result_img, edgemap, "RIGHT")
display_image(result_img, title="", cmap=None)
#save_image(file_name+"_sorting_narrow_down"+".png", result_img)
