def crop_augmentor(train_x, train_y, num_per_class, aug_with='pos')
Generate new images by picking pairs of images and randomly replace region of one, quarter or half, by the same region of the other.
x_aug,y_aug = crop_augmentor(train_x,train_y,1000, aug_with='pos')train_x: numpy array with the names of images files.
train_y: numpy array with labels of the images.
num_per_class: the number of generated images per class.
aug_with: if `pos` the pairs are of the same class. If `neg` the pairs of different classes. If `le` the pairs are of classes that are less or equal to the generated label.
train_x: numpy array with the base images names and the generated images names.
train_y: numpy array with labels of the base images and the generated images.
def background_noiser(img, gray_threshold=10)
Adds noise to regions of the image with a value less than the threshold.
img: numpy array of the image. gray_threshold: gray threshold.
img: the image with noise over the regions that have gray values less than gray_threshold.
def circle_centering(image, circle_detection_method='moments', scale_value=1, aspect_ratio=1.6, width=224, gray_threshold=10)
Applies Affine Transformation to move and scale the circle to the center of the image (diabetic retinopathy)
image: numpy array
circle_detection_method: moments, enclosing_circle or max_dim. moments to get the centroid of the binary image obtained by binary comparison of the gray image with gray_threshold and the radius of a circle having area equivalent to that of the binary image. enclosing_circle detects the enclosing circle of the largest contour and uses it's center and radius. max_dim uses the max dimension of the image as the raduis and the center of the image as the center
scale_value: float value by which the detected circle is scaled.
aspect_ratio: aspect ratio for crop.
width: the width of the resulting image.
gray_threshold: int used to separate the eye from the background.
Output
image: numpy array contains the circle detected by the method in the center and having a raduis of the width.
def edger(img, gray_threshold=15, kernel_size=5, dilate_itr=1)
def pyramid_scale_down(img, width=300)
def veins_spots_highlighter(src, ddepth=3, kernel_size=3)
def apply_preprocess(names, src_path, dst_path=None, extension='png', preprocessing_function=[], preprocessing_params=[], write=False, prog_bar_disable=False)
Applys list of functions to list of images specified by name and path
circle_centering_params = {
'circle_detection_method':'enclosing_circle',
'scale_value' : scale_value,
'aspect_ratio' : aspect_ratio,
'width' : image_size,
'gray_threshold' : 12
}
preprocess_sequence = [circle_centering]
preprocess_params = [circle_centering_params]
lister = np.random.choice(df_train['id_code'],25)
imgs = apply_preprocess(lister,src_path=preprocessed_path+'train/', preprocessing_function=[preprocess_sequence], preprocessing_params=[preprocess_params])
for i in range(25):
plt.subplot(5,5,i+1)
plt.imshow(imgs[i])names: list of image names
src_path: source path
dst_path: destination path
extension: image extension
preprocessing_function: list of functions to apply in order on the image
preprocessing_params: list of dictionaries of the same length as preprocessing_function that contain the parameter for each function in order.
write: if True the resulting images will be written in the dst_path. If False the resultingimages are returned.
prog_bar_disable: if False the function will show the progress bar. If True the function will not show the progress bar.
Returns the resuling images if write==False and returns and empty list if write==True.
def balanced_valid_set_splitter(X, y, classes=None, valid_n_per_class=30, shuffle=False, debug=False)
Splits the dataset to a training set and validation set. The resulting validation set is balanced.
X: numpy array of the examples.
y: numpy array of the labels.
classes: list of labels the dataset. If None is provided,classes will be set as follows classes = np.unique(y, axis=0). valid_n_per_class: Number of examples per class in the validation set. shuffle: shuffles the dataset if shuffle==True. debug: displays histograms of the training set and the validation set if set to True`.
train_x: numpy array of the training examples. train_y: numpy array of the training labels. valid_x: numpy array of the validation examples. valid_y: numpy array of the validation labels.
def classes_decoder(y, method='max')
Decodes the results of classification problem.
y: numpy array of the prediction from a classification NN of shape (number of examples, number of classes).
method: if max the examples are classified based on max value. If highest_true the examples are classified based on the highest true class as if the examples have multilabels.
numpy array with the classification.
def classes_encoder(y, n_classes, method='one')
Encodes the labels of the dataset.
y: numpy array of the classes.
n_classes: number of classes in y.
method: if one the examples are encoded in one-hot encoding. If all_lower_ones the examples are encoded as a multilabel classes that have all lower classes as labels. If pairs the examples are encoded as a multilabel classes that have the labels of it's class and the next lower class.
y_coded: numpy array contains the encoded ys.