import sys
sys.path.insert(0, __folder_params.home)
import utils
from skimage import exposure
import cv2
def processHistogram(img_path, radius = 15):
""" Extracts histogram for each color channel. """
img = cv2.imread(img_path)
final_path = utils.adress_file(img_path, "ColorHistogram")
# http://scikit-image.org/docs/dev/api/skimage.exposure.html#skimage.exposure.histogram
R, G, B = img[:, :, 2], img[:, :, 1], img[:, :, 0]
Rhistogram = exposure.histogram(R)
Ghistogram = exposure.histogram(G)
Bhistogram = exposure.histogram(B)
with open(final_path, 'w+') as f:
for histogram in (Rhistogram, Ghistogram, Bhistogram):
values, bins = histogram
for v, b in zip(values, bins):
f.write("%d:%d " % (b, v))
f.write("\n")
if __name__ == "__main__":
utils.FeatureExtractor("image", processHistogram).run()
#a = FeatureExtractor("image", processHistogram)
#print(next(a.data_iterator))
#processHistogram(next(a.data_iterator))
input_shape,
mode='constant',
clip=True,
preserve_range=True)
X[i, :, :, :] = img
out = model.predict(X)
for i, pred in enumerate(out):
with open(final_paths[i], 'w+') as f:
for i, v in enumerate(pred):
if v > 0:
f.write("%d:%e " % (i, v))
f.write("\n")
if __name__ == "__main__":
#FeatureExtractor("image", processInceptionV3).run()
a = utils.FeatureExtractor("image", processDnn)
images_paths = [i for i in a.data_iterator]
batch = 32
batches = []
for i in range(len(images_paths) // 32):
batches.append(images_paths[i * 32:(i + 1) * 32])
batches.append(images_paths[(i + 1) * 32:])
model = get_inceptionV3()
for img_batch in tqdm(batches, total=len(batches)):
processDnn(img_batch, model)
def perform_fft(self, window_size: float): # filtered_data = self.visible_filtered_data() # length = pow(2, utils.closest_power_of_two(filtered_data.shape[0])) # frequencies = np.linspace(0, BoardShim.get_sampling_rate(BOARD_ID) / 2, int(length / 2 + 1)) # return frequencies, DataFilter.perform_fft(filtered_data[:length], WindowFunctions.NO_WINDOW.value) return utils.FeatureExtractor(self.visible_filtered_data(), global_config.SAMPLING_RATE).fft(window_size)
import cv2
def processHog(img_path, pixels_per_cell=(32, 32), cells_per_block=(1, 1)):
final_path = utils.adress_file(img_path, "Hog")
# http://scikit-image.org/docs/dev/api/skimage.feature.html#skimage.feature.hog
img = cv2.imread(img_path)
im_gray = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)
fd = hog(im_gray,
block_norm='L2-Hys',
cells_per_block=cells_per_block,
pixels_per_cell=pixels_per_cell)
# Calculate the histogram
x = itemfreq(fd.ravel())
# Normalize the histogram
hist = x[:, 1] / sum(x[:, 1])
with open(final_path, 'w+') as f:
for v in hist:
f.write("%.8f " % (v))
f.write("\n")
if __name__ == "__main__":
utils.FeatureExtractor("image", processHog).run()
#a = FeatureExtractor("image", processHog)
#print(next(a.data_iterator))
#processHog(next(a.data_iterator))
#processHog(next(a.data_iterator))