Module monk.gluon.transforms.transforms

Functions

def transform_center_crop(system_dict, input_size, train, val, test, retrieve=False)

Apply Center Cropping transformation

Args

system_dict : dict

System dictionary storing experiment state and set variables

input_size : int, list

Crop size

train : bool

If True, transform applied to training data

val : bool

If True, transform applied to validation data

test : bool

If True, transform applied to testing/inferencing data

Returns

dict

updated system dict

Expand source code

def transform_center_crop(system_dict, input_size, train, val, test, retrieve=False):
    '''
    Apply Center Cropping transformation

    Args:
        system_dict (dict): System dictionary storing experiment state and set variables
        input_size (int, list): Crop size
        train (bool): If True, transform applied to training data
        val (bool): If True, transform applied to validation data
        test (bool): If True, transform applied to testing/inferencing data

    Returns:
        dict: updated system dict
    '''
    tmp = {};
    tmp["CenterCrop"] = {};
    tmp["CenterCrop"]["input_size"] = input_size;

    if(train):
        if(not retrieve):
            system_dict["dataset"]["transforms"]["train"].append(tmp);
        system_dict["local"]["transforms_train"].append(transforms.CenterCrop(input_size));
    if(val):
        if(not retrieve):
            system_dict["dataset"]["transforms"]["val"].append(tmp);
        system_dict["local"]["transforms_val"].append(transforms.CenterCrop(input_size));
    if(test):
        if(not retrieve):
            system_dict["dataset"]["transforms"]["test"].append(tmp);
        system_dict["local"]["transforms_test"].append(transforms.CenterCrop(input_size));

    return system_dict;

def transform_color_jitter(system_dict, brightness, contrast, saturation, hue, train, val, test, retrieve=False)

Apply Color jittering transformations

Args

system_dict : dict

System dictionary storing experiment state and set variables

brightness : float

Levels to jitter brightness. 0 - min 1 - max

contrast : float

Levels to jitter contrast. 0 - min 1 - max

saturation : float

Levels to jitter saturation. 0 - min 1 - max

hue : float

Levels to jitter hue. 0 - min 1 - max

train : bool

If True, transform applied to training data

val : bool

If True, transform applied to validation data

test : bool

If True, transform applied to testing/inferencing data

Returns

dict

updated system dict

Expand source code

def transform_color_jitter(system_dict, brightness, contrast, saturation, hue, train, val, test, retrieve=False):
    '''
    Apply Color jittering transformations

    Args:
        system_dict (dict): System dictionary storing experiment state and set variables
        brightness (float): Levels to jitter brightness.
                                    0 - min
                                    1 - max
        contrast (float): Levels to jitter contrast.
                                    0 - min
                                    1 - max
        saturation (float): Levels to jitter saturation.
                                    0 - min
                                    1 - max
        hue (float): Levels to jitter hue.
                                    0 - min
                                    1 - max
        train (bool): If True, transform applied to training data
        val (bool): If True, transform applied to validation data
        test (bool): If True, transform applied to testing/inferencing data

    Returns:
        dict: updated system dict
    '''
    tmp = {};
    tmp["ColorJitter"] = {};
    tmp["ColorJitter"]["brightness"] = brightness;
    tmp["ColorJitter"]["contrast"] = contrast;
    tmp["ColorJitter"]["saturation"] = saturation;
    tmp["ColorJitter"]["hue"] = hue;

    if(train):
        if(not retrieve):
            system_dict["dataset"]["transforms"]["train"].append(tmp);
        system_dict["local"]["transforms_train"].append(transforms.ColorJitter(brightness=brightness, contrast=contrast, saturation=saturation, hue=hue));
    if(val):
        if(not retrieve):
            system_dict["dataset"]["transforms"]["val"].append(tmp);
        system_dict["local"]["transforms_val"].append(transforms.ColorJitter(brightness=brightness, contrast=contrast, saturation=saturation, hue=hue));
    if(test):
        if(not retrieve):
            system_dict["dataset"]["transforms"]["test"].append(tmp);
        system_dict["local"]["transforms_test"].append(transforms.ColorJitter(brightness=brightness, contrast=contrast, saturation=saturation, hue=hue));


    return system_dict;

def transform_normalize(system_dict, mean, std, train, val, test, retrieve=False)

Apply mean subtraction and standard normalization

Args

system_dict : dict

System dictionary storing experiment state and set variables

mean : float, list

Mean value for subtraction

std : float, list

Normalization factor

train : bool

If True, transform applied to training data

val : bool

If True, transform applied to validation data

test : bool

If True, transform applied to testing/inferencing data

Returns

dict

updated system dict

Expand source code

def transform_normalize(system_dict, mean, std, train, val, test, retrieve=False):
    '''
    Apply mean subtraction and standard normalization

    Args:
        system_dict (dict): System dictionary storing experiment state and set variables
        mean (float, list): Mean value for subtraction
        std (float, list): Normalization factor
        train (bool): If True, transform applied to training data
        val (bool): If True, transform applied to validation data
        test (bool): If True, transform applied to testing/inferencing data

    Returns:
        dict: updated system dict
    '''
    tmp = {};
    tmp["Normalize"] = {};
    tmp["Normalize"]["mean"] = mean;
    tmp["Normalize"]["std"] = std;
    system_dict["local"]["normalize"] = True;
    input_size = system_dict["dataset"]["params"]["input_size"];

    if(type(system_dict["dataset"]["params"]["input_size"]) == tuple or type(system_dict["dataset"]["params"]["input_size"]) == list):
        h = system_dict["dataset"]["params"]["input_size"][0];
        w = system_dict["dataset"]["params"]["input_size"][1];
    else:
        h = system_dict["dataset"]["params"]["input_size"];
        w = system_dict["dataset"]["params"]["input_size"];

    if(train):
        if(not retrieve):
            system_dict["dataset"]["transforms"]["train"].append(tmp);
        system_dict["local"]["transforms_train"].append(transforms.Resize(size=(w, h)));
        system_dict["local"]["transforms_train"].append(transforms.ToTensor())
        system_dict["local"]["transforms_train"].append(transforms.Normalize(mean=mean, std=std));
    if(val):
        if(not retrieve):
            system_dict["dataset"]["transforms"]["val"].append(tmp);
        system_dict["local"]["transforms_val"].append(transforms.Resize(size=(w, h)));
        system_dict["local"]["transforms_val"].append(transforms.ToTensor())
        system_dict["local"]["transforms_val"].append(transforms.Normalize(mean=mean, std=std));
    if(test):
        if(not retrieve):
            system_dict["dataset"]["transforms"]["test"].append(tmp);
        system_dict["local"]["transforms_test"].append(transforms.Resize(size=(w, h)));
        system_dict["local"]["transforms_test"].append(transforms.ToTensor())
        system_dict["local"]["transforms_test"].append(transforms.Normalize(mean=mean, std=std));

    return system_dict;

def transform_random_horizontal_flip(system_dict, probability, train, val, test, retrieve=False)

Apply random horizontal flip transformations

Args

system_dict : dict

System dictionary storing experiment state and set variables

probability : float

Probability of flipping the input image

train : bool

If True, transform applied to training data

val : bool

If True, transform applied to validation data

test : bool

If True, transform applied to testing/inferencing data

Returns

dict

updated system dict

Expand source code

def transform_random_horizontal_flip(system_dict, probability, train, val, test, retrieve=False):
    '''
    Apply random horizontal flip transformations

    Args:
        system_dict (dict): System dictionary storing experiment state and set variables
        probability (float): Probability of flipping the input image
        train (bool): If True, transform applied to training data
        val (bool): If True, transform applied to validation data
        test (bool): If True, transform applied to testing/inferencing data

    Returns:
        dict: updated system dict
    '''
    tmp = {};
    tmp["RandomHorizontalFlip"] = {};
    tmp["RandomHorizontalFlip"]["p"] = probability;

    if(train):
        if(not retrieve):
            system_dict["dataset"]["transforms"]["train"].append(tmp);
        system_dict["local"]["transforms_train"].append(transforms.RandomFlipLeftRight());
    if(val):
        if(not retrieve):
            system_dict["dataset"]["transforms"]["val"].append(tmp);
        system_dict["local"]["transforms_val"].append(transforms.RandomFlipLeftRight());
    if(test):
        if(not retrieve):
            system_dict["dataset"]["transforms"]["test"].append(tmp);
        system_dict["local"]["transforms_test"].append(transforms.RandomFlipLeftRight());


    return system_dict;

def transform_random_lighting(system_dict, alpha, train, val, test, retrieve=False)

Apply random lighting transformations

Args

system_dict : dict

System dictionary storing experiment state and set variables

alpha : float

1.0 - Randomly increase overall lighting intensity <1.0 - Randomly decrease overall lighting intensity

train : bool

If True, transform applied to training data

val : bool

If True, transform applied to validation data

test : bool

If True, transform applied to testing/inferencing data

Returns

dict

updated system dict

Expand source code

def transform_random_lighting(system_dict, alpha, train, val, test, retrieve=False):
    '''
    Apply random lighting transformations

    Args:
        system_dict (dict): System dictionary storing experiment state and set variables
        alpha (float): > 1.0 - Randomly increase overall lighting intensity
                        <1.0 - Randomly decrease overall lighting intensity
        train (bool): If True, transform applied to training data
        val (bool): If True, transform applied to validation data
        test (bool): If True, transform applied to testing/inferencing data

    Returns:
        dict: updated system dict
    '''
    tmp = {};
    tmp["RandomLighting"] = {};
    tmp["RandomLighting"]["alpha"] = alpha;

    if(train):
        if(not retrieve):
            system_dict["dataset"]["transforms"]["train"].append(tmp);
        system_dict["local"]["transforms_train"].append(transforms.RandomLighting(alpha));
    if(val):
        if(not retrieve):
            system_dict["dataset"]["transforms"]["val"].append(tmp);
        system_dict["local"]["transforms_val"].append(transforms.RandomLighting(alpha));
    if(test):
        if(not retrieve):
            system_dict["dataset"]["transforms"]["test"].append(tmp);
        system_dict["local"]["transforms_test"].append(transforms.RandomLighting(alpha));


    return system_dict;

def transform_random_resized_crop(system_dict, input_size, scale, ratio, train, val, test, retrieve=False)

Apply Random Resized Cropping transformation

Args

system_dict : dict

System dictionary storing experiment state and set variables

input_size : int, list

Crop size

scale : float, tuple

scaling ratio limits; for maximum and minimum random scaling

ratio : float, tuple

aspect ratio limits; for maximum and minmum changes to aspect ratios

train : bool

If True, transform applied to training data

val : bool

If True, transform applied to validation data

test : bool

If True, transform applied to testing/inferencing data

Returns

dict

updated system dict

Expand source code

def transform_random_resized_crop(system_dict, input_size, scale, ratio, train, val, test, retrieve=False):
    '''
    Apply Random Resized Cropping transformation

    Args:
        system_dict (dict): System dictionary storing experiment state and set variables
        input_size (int, list): Crop size
        scale (float, tuple): scaling ratio limits; for maximum and minimum random scaling
        ratio (float, tuple): aspect ratio limits; for maximum and minmum changes to aspect ratios 
        train (bool): If True, transform applied to training data
        val (bool): If True, transform applied to validation data
        test (bool): If True, transform applied to testing/inferencing data

    Returns:
        dict: updated system dict
    '''
    tmp = {};
    tmp["RandomResizedCrop"] = {};
    tmp["RandomResizedCrop"]["input_size"] = input_size;
    tmp["RandomResizedCrop"]["scale"] = scale;
    tmp["RandomResizedCrop"]["ratio"] = ratio;


    if(train):
        if(not retrieve):
            system_dict["dataset"]["transforms"]["train"].append(tmp);
        system_dict["local"]["transforms_train"].append(transforms.RandomResizedCrop(size=input_size, scale=scale, ratio=ratio));
    if(val):
        if(not retrieve):
            system_dict["dataset"]["transforms"]["val"].append(tmp);
        system_dict["local"]["transforms_val"].append(transforms.RandomResizedCrop(size=input_size, scale=scale, ratio=ratio));
    if(test):
        if(not retrieve):
            system_dict["dataset"]["transforms"]["test"].append(tmp);
        system_dict["local"]["transforms_test"].append(transforms.RandomResizedCrop(size=input_size, scale=scale, ratio=ratio));


    return system_dict;

def transform_random_vertical_flip(system_dict, probability, train, val, test, retrieve=False)

Apply random vertical flip transformations

Args

system_dict : dict

System dictionary storing experiment state and set variables

probability : float

Probability of flipping the input image

train : bool

If True, transform applied to training data

val : bool

If True, transform applied to validation data

test : bool

If True, transform applied to testing/inferencing data

Returns

dict

updated system dict

Expand source code

def transform_random_vertical_flip(system_dict, probability, train, val, test, retrieve=False):
    '''
    Apply random vertical flip transformations

    Args:
        system_dict (dict): System dictionary storing experiment state and set variables
        probability (float): Probability of flipping the input image
        train (bool): If True, transform applied to training data
        val (bool): If True, transform applied to validation data
        test (bool): If True, transform applied to testing/inferencing data

    Returns:
        dict: updated system dict
    '''
    tmp = {};
    tmp["RandomVerticalFlip"] = {};
    tmp["RandomVerticalFlip"]["p"] = probability;

    if(train):
        if(not retrieve):
            system_dict["dataset"]["transforms"]["train"].append(tmp);
        system_dict["local"]["transforms_train"].append(transforms.RandomFlipTopBottom());
    if(val):
        if(not retrieve):
            system_dict["dataset"]["transforms"]["val"].append(tmp);
        system_dict["local"]["transforms_val"].append(transforms.RandomFlipTopBottom());
    if(test):
        if(not retrieve):
            system_dict["dataset"]["transforms"]["test"].append(tmp);
        system_dict["local"]["transforms_test"].append(transforms.RandomFlipTopBottom());

    return system_dict;

def transform_resize(system_dict, input_size, train, val, test, retrieve=False)

Apply standard resizing

Args

system_dict : dict

System dictionary storing experiment state and set variables

input_size : int, list

expected final size

train : bool

If True, transform applied to training data

val : bool

If True, transform applied to validation data

test : bool

If True, transform applied to testing/inferencing data

Returns

dict

updated system dict

Expand source code

def transform_resize(system_dict, input_size, train, val, test, retrieve=False):
    '''
    Apply standard resizing

    Args:
        system_dict (dict): System dictionary storing experiment state and set variables
        input_size (int, list): expected final size
        train (bool): If True, transform applied to training data
        val (bool): If True, transform applied to validation data
        test (bool): If True, transform applied to testing/inferencing data

    Returns:
        dict: updated system dict
    '''
    tmp = {};
    tmp["Resize"] = {};
    tmp["Resize"]["input_size"] = input_size;

    if(type(input_size) == tuple or type(input_size) == list):
        h = input_size[0];
        w = input_size[1];
    else:
        h = input_size;
        w = input_size;

    if(train):
        if(not retrieve):
            system_dict["dataset"]["transforms"]["train"].append(tmp);
        system_dict["local"]["transforms_train"].append(transforms.Resize(size=(w, h)));
    if(val):
        if(not retrieve):
            system_dict["dataset"]["transforms"]["val"].append(tmp);
        system_dict["local"]["transforms_val"].append(transforms.Resize(size=(w, h)));
    if(test):
        if(not retrieve):
            system_dict["dataset"]["transforms"]["test"].append(tmp);
        system_dict["local"]["transforms_test"].append(transforms.Resize(size=(w, h)));

    return system_dict;