transforms.py

# coding: utf8
# Copyright (c) 2020 PaddlePaddle Authors. All Rights Reserved.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
#    http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.

import random
from collections import OrderedDict

import numpy as np
from PIL import Image
import cv2

from .functional import *


class Compose:
    def __init__(self, transforms, to_rgb=True):
        if not isinstance(transforms, list):
            raise TypeError('The transforms must be a list!')
        if len(transforms) < 1:
            raise ValueError('The length of transforms ' + \
                            'must be equal or larger than 1!')
        self.transforms = transforms
        self.to_rgb = to_rgb

    def __call__(self, im, im_info=None, label=None):
        if im_info is None:
            im_info = list()
        if isinstance(im, str):
            im = cv2.imread(im).astype('float32')
        if isinstance(label, str):
            label = np.asarray(Image.open(label))
        if im is None:
            raise ValueError('Can\'t read The image file {}!'.format(im))
        if self.to_rgb:
            im = cv2.cvtColor(im, cv2.COLOR_BGR2RGB)

        for op in self.transforms:
            outputs = op(im, im_info, label)
            im = outputs[0]
            if len(outputs) >= 2:
                im_info = outputs[1]
            if len(outputs) == 3:
                label = outputs[2]
        im = permute(im)
        if len(outputs) == 3:
            label = label[np.newaxis, :, :]
        return (im, im_info, label)


class RandomHorizontalFlip:
    def __init__(self, prob=0.5):
        self.prob = prob

    def __call__(self, im, im_info=None, label=None):
        if random.random() < self.prob:
            im = horizontal_flip(im)
            if label is not None:
                label = horizontal_flip(label)
        if label is None:
            return (im, im_info)
        else:
            return (im, im_info, label)


class RandomVerticalFlip:
    def __init__(self, prob=0.1):
        self.prob = prob

    def __call__(self, im, im_info=None, label=None):
        if random.random() < self.prob:
            im = vertical_flip(im)
            if label is not None:
                label = vertical_flip(label)
        if label is None:
            return (im, im_info)
        else:
            return (im, im_info, label)


class Resize:
    # The interpolation mode
    interp_dict = {
        'NEAREST': cv2.INTER_NEAREST,
        'LINEAR': cv2.INTER_LINEAR,
        'CUBIC': cv2.INTER_CUBIC,
        'AREA': cv2.INTER_AREA,
        'LANCZOS4': cv2.INTER_LANCZOS4
    }

    def __init__(self, target_size=512, interp='LINEAR'):
        self.interp = interp
        if not (interp == "RANDOM" or interp in self.interp_dict):
            raise ValueError("interp should be one of {}".format(
                self.interp_dict.keys()))
        if isinstance(target_size, list) or isinstance(target_size, tuple):
            if len(target_size) != 2:
                raise TypeError(
                    'when target is list or tuple, it should include 2 elements, but it is {}'
                    .format(target_size))
        elif not isinstance(target_size, int):
            raise TypeError(
                "Type of target_size is invalid. Must be Integer or List or tuple, now is {}"
                .format(type(target_size)))

        self.target_size = target_size

    def __call__(self, im, im_info=None, label=None):
        if im_info is None:
            im_info = list()
        im_info.append(('resize', im.shape[:2]))
        if not isinstance(im, np.ndarray):
            raise TypeError("Resize: image type is not numpy.")
        if len(im.shape) != 3:
            raise ValueError('Resize: image is not 3-dimensional.')
        if self.interp == "RANDOM":
            interp = random.choice(list(self.interp_dict.keys()))
        else:
            interp = self.interp
        im = resize(im, self.target_size, self.interp_dict[interp])
        if label is not None:
            label = resize(label, self.target_size, cv2.INTER_NEAREST)

        if label is None:
            return (im, im_info)
        else:
            return (im, im_info, label)


class ResizeByLong:
    def __init__(self, long_size):
        self.long_size = long_size

    def __call__(self, im, im_info=None, label=None):
        if im_info is None:
            im_info = list()

        im_info.append(('resize', im.shape[:2]))
        im = resize_long(im, self.long_size)
        if label is not None:
            label = resize_long(label, self.long_size, cv2.INTER_NEAREST)

        if label is None:
            return (im, im_info)
        else:
            return (im, im_info, label)


class ResizeRangeScaling:
    def __init__(self, min_value=400, max_value=600):
        if min_value > max_value:
            raise ValueError('min_value must be less than max_value, '
                             'but they are {} and {}.'.format(
                                 min_value, max_value))
        self.min_value = min_value
        self.max_value = max_value

    def __call__(self, im, im_info=None, label=None):
        if self.min_value == self.max_value:
            random_size = self.max_value
        else:
            random_size = int(
                np.random.uniform(self.min_value, self.max_value) + 0.5)
        im = resize_long(im, random_size, cv2.INTER_LINEAR)
        if label is not None:
            label = resize_long(label, random_size, cv2.INTER_NEAREST)

        if label is None:
            return (im, im_info)
        else:
            return (im, im_info, label)


class ResizeStepScaling:
    def __init__(self,
                 min_scale_factor=0.75,
                 max_scale_factor=1.25,
                 scale_step_size=0.25):
        if min_scale_factor > max_scale_factor:
            raise ValueError(
                'min_scale_factor must be less than max_scale_factor, '
                'but they are {} and {}.'.format(min_scale_factor,
                                                 max_scale_factor))
        self.min_scale_factor = min_scale_factor
        self.max_scale_factor = max_scale_factor
        self.scale_step_size = scale_step_size

    def __call__(self, im, im_info=None, label=None):
        if self.min_scale_factor == self.max_scale_factor:
            scale_factor = self.min_scale_factor

        elif self.scale_step_size == 0:
            scale_factor = np.random.uniform(self.min_scale_factor,
                                             self.max_scale_factor)

        else:
            num_steps = int((self.max_scale_factor - self.min_scale_factor) /
                            self.scale_step_size + 1)
            scale_factors = np.linspace(self.min_scale_factor,
                                        self.max_scale_factor,
                                        num_steps).tolist()
            np.random.shuffle(scale_factors)
            scale_factor = scale_factors[0]
        w = int(round(scale_factor * im.shape[1]))
        h = int(round(scale_factor * im.shape[0]))

        im = resize(im, (w, h), cv2.INTER_LINEAR)
        if label is not None:
            label = resize(label, (w, h), cv2.INTER_NEAREST)

        if label is None:
            return (im, im_info)
        else:
            return (im, im_info, label)


class Normalize:
    def __init__(self, mean=[0.5, 0.5, 0.5], std=[0.5, 0.5, 0.5]):
        self.mean = mean
        self.std = std
        if not (isinstance(self.mean, list) and isinstance(self.std, list)):
            raise ValueError("{}: input type is invalid.".format(self))
        from functools import reduce
        if reduce(lambda x, y: x * y, self.std) == 0:
            raise ValueError('{}: std is invalid!'.format(self))

    def __call__(self, im, im_info=None, label=None):
        mean = np.array(self.mean)[np.newaxis, np.newaxis, :]
        std = np.array(self.std)[np.newaxis, np.newaxis, :]
        im = normalize(im, mean, std)

        if label is None:
            return (im, im_info)
        else:
            return (im, im_info, label)


class Padding:
    def __init__(self,
                 target_size,
                 im_padding_value=[127.5, 127.5, 127.5],
                 label_padding_value=255):
        if isinstance(target_size, list) or isinstance(target_size, tuple):
            if len(target_size) != 2:
                raise ValueError(
                    'when target is list or tuple, it should include 2 elements, but it is {}'
                    .format(target_size))
        elif not isinstance(target_size, int):
            raise TypeError(
                "Type of target_size is invalid. Must be Integer or List or tuple, now is {}"
                .format(type(target_size)))
        self.target_size = target_size
        self.im_padding_value = im_padding_value
        self.label_padding_value = label_padding_value

    def __call__(self, im, im_info=None, label=None):
        if im_info is None:
            im_info = list()
        im_info.append(('padding', im.shape[:2]))

        im_height, im_width = im.shape[0], im.shape[1]
        if isinstance(self.target_size, int):
            target_height = self.target_size
            target_width = self.target_size
        else:
            target_height = self.target_size[1]
            target_width = self.target_size[0]
        pad_height = target_height - im_height
        pad_width = target_width - im_width
        if pad_height < 0 or pad_width < 0:
            raise ValueError(
                'the size of image should be less than target_size, but the size of image ({}, {}), is larger than target_size ({}, {})'
                .format(im_width, im_height, target_width, target_height))
        else:
            im = cv2.copyMakeBorder(
                im,
                0,
                pad_height,
                0,
                pad_width,
                cv2.BORDER_CONSTANT,
                value=self.im_padding_value)
            if label is not None:
                label = cv2.copyMakeBorder(
                    label,
                    0,
                    pad_height,
                    0,
                    pad_width,
                    cv2.BORDER_CONSTANT,
                    value=self.label_padding_value)
        if label is None:
            return (im, im_info)
        else:
            return (im, im_info, label)


class RandomPaddingCrop:
    def __init__(self,
                 crop_size=512,
                 im_padding_value=[127.5, 127.5, 127.5],
                 label_padding_value=255):
        if isinstance(crop_size, list) or isinstance(crop_size, tuple):
            if len(crop_size) != 2:
                raise ValueError(
                    'when crop_size is list or tuple, it should include 2 elements, but it is {}'
                    .format(crop_size))
        elif not isinstance(crop_size, int):
            raise TypeError(
                "Type of crop_size is invalid. Must be Integer or List or tuple, now is {}"
                .format(type(crop_size)))
        self.crop_size = crop_size
        self.im_padding_value = im_padding_value
        self.label_padding_value = label_padding_value

    def __call__(self, im, im_info=None, label=None):
        if isinstance(self.crop_size, int):
            crop_width = self.crop_size
            crop_height = self.crop_size
        else:
            crop_width = self.crop_size[0]
            crop_height = self.crop_size[1]

        img_height = im.shape[0]
        img_width = im.shape[1]

        if img_height == crop_height and img_width == crop_width:
            if label is None:
                return (im, im_info)
            else:
                return (im, im_info, label)
        else:
            pad_height = max(crop_height - img_height, 0)
            pad_width = max(crop_width - img_width, 0)
            if (pad_height > 0 or pad_width > 0):
                im = cv2.copyMakeBorder(
                    im,
                    0,
                    pad_height,
                    0,
                    pad_width,
                    cv2.BORDER_CONSTANT,
                    value=self.im_padding_value)
                if label is not None:
                    label = cv2.copyMakeBorder(
                        label,
                        0,
                        pad_height,
                        0,
                        pad_width,
                        cv2.BORDER_CONSTANT,
                        value=self.label_padding_value)
                img_height = im.shape[0]
                img_width = im.shape[1]

            if crop_height > 0 and crop_width > 0:
                h_off = np.random.randint(img_height - crop_height + 1)
                w_off = np.random.randint(img_width - crop_width + 1)

                im = im[h_off:(crop_height + h_off), w_off:(
                    w_off + crop_width), :]
                if label is not None:
                    label = label[h_off:(crop_height + h_off), w_off:(
                        w_off + crop_width)]
        if label is None:
            return (im, im_info)
        else:
            return (im, im_info, label)


class RandomBlur:
    def __init__(self, prob=0.1):
        self.prob = prob

    def __call__(self, im, im_info=None, label=None):
        if self.prob <= 0:
            n = 0
        elif self.prob >= 1:
            n = 1
        else:
            n = int(1.0 / self.prob)
        if n > 0:
            if np.random.randint(0, n) == 0:
                radius = np.random.randint(3, 10)
                if radius % 2 != 1:
                    radius = radius + 1
                if radius > 9:
                    radius = 9
                im = cv2.GaussianBlur(im, (radius, radius), 0, 0)

        if label is None:
            return (im, im_info)
        else:
            return (im, im_info, label)


class RandomRotation:
    def __init__(self,
                 max_rotation=15,
                 im_padding_value=[127.5, 127.5, 127.5],
                 label_padding_value=255):
        self.max_rotation = max_rotation
        self.im_padding_value = im_padding_value
        self.label_padding_value = label_padding_value

    def __call__(self, im, im_info=None, label=None):
        if self.max_rotation > 0:
            (h, w) = im.shape[:2]
            do_rotation = np.random.uniform(-self.max_rotation,
                                            self.max_rotation)
            pc = (w // 2, h // 2)
            r = cv2.getRotationMatrix2D(pc, do_rotation, 1.0)
            cos = np.abs(r[0, 0])
            sin = np.abs(r[0, 1])

            nw = int((h * sin) + (w * cos))
            nh = int((h * cos) + (w * sin))

            (cx, cy) = pc
            r[0, 2] += (nw / 2) - cx
            r[1, 2] += (nh / 2) - cy
            dsize = (nw, nh)
            im = cv2.warpAffine(
                im,
                r,
                dsize=dsize,
                flags=cv2.INTER_LINEAR,
                borderMode=cv2.BORDER_CONSTANT,
                borderValue=self.im_padding_value)
            label = cv2.warpAffine(
                label,
                r,
                dsize=dsize,
                flags=cv2.INTER_NEAREST,
                borderMode=cv2.BORDER_CONSTANT,
                borderValue=self.label_padding_value)

        if label is None:
            return (im, im_info)
        else:
            return (im, im_info, label)


class RandomScaleAspect:
    def __init__(self, min_scale=0.5, aspect_ratio=0.33):
        self.min_scale = min_scale
        self.aspect_ratio = aspect_ratio

    def __call__(self, im, im_info=None, label=None):
        if self.min_scale != 0 and self.aspect_ratio != 0:
            img_height = im.shape[0]
            img_width = im.shape[1]
            for i in range(0, 10):
                area = img_height * img_width
                target_area = area * np.random.uniform(self.min_scale, 1.0)
                aspectRatio = np.random.uniform(self.aspect_ratio,
                                                1.0 / self.aspect_ratio)

                dw = int(np.sqrt(target_area * 1.0 * aspectRatio))
                dh = int(np.sqrt(target_area * 1.0 / aspectRatio))
                if (np.random.randint(10) < 5):
                    tmp = dw
                    dw = dh
                    dh = tmp

                if (dh < img_height and dw < img_width):
                    h1 = np.random.randint(0, img_height - dh)
                    w1 = np.random.randint(0, img_width - dw)

                    im = im[h1:(h1 + dh), w1:(w1 + dw), :]
                    label = label[h1:(h1 + dh), w1:(w1 + dw)]
                    im = cv2.resize(
                        im, (img_width, img_height),
                        interpolation=cv2.INTER_LINEAR)
                    label = cv2.resize(
                        label, (img_width, img_height),
                        interpolation=cv2.INTER_NEAREST)
                    break
        if label is None:
            return (im, im_info)
        else:
            return (im, im_info, label)


class RandomDistort:
    def __init__(self,
                 brightness_range=0.5,
                 brightness_prob=0.5,
                 contrast_range=0.5,
                 contrast_prob=0.5,
                 saturation_range=0.5,
                 saturation_prob=0.5,
                 hue_range=18,
                 hue_prob=0.5):
        self.brightness_range = brightness_range
        self.brightness_prob = brightness_prob
        self.contrast_range = contrast_range
        self.contrast_prob = contrast_prob
        self.saturation_range = saturation_range
        self.saturation_prob = saturation_prob
        self.hue_range = hue_range
        self.hue_prob = hue_prob

    def __call__(self, im, im_info=None, label=None):
        brightness_lower = 1 - self.brightness_range
        brightness_upper = 1 + self.brightness_range
        contrast_lower = 1 - self.contrast_range
        contrast_upper = 1 + self.contrast_range
        saturation_lower = 1 - self.saturation_range
        saturation_upper = 1 + self.saturation_range
        hue_lower = -self.hue_range
        hue_upper = self.hue_range
        ops = [brightness, contrast, saturation, hue]
        random.shuffle(ops)
        params_dict = {
            'brightness': {
                'brightness_lower': brightness_lower,
                'brightness_upper': brightness_upper
            },
            'contrast': {
                'contrast_lower': contrast_lower,
                'contrast_upper': contrast_upper
            },
            'saturation': {
                'saturation_lower': saturation_lower,
                'saturation_upper': saturation_upper
            },
            'hue': {
                'hue_lower': hue_lower,
                'hue_upper': hue_upper
            }
        }
        prob_dict = {
            'brightness': self.brightness_prob,
            'contrast': self.contrast_prob,
            'saturation': self.saturation_prob,
            'hue': self.hue_prob
        }
        im = im.astype('uint8')
        im = Image.fromarray(im)
        for id in range(4):
            params = params_dict[ops[id].__name__]
            prob = prob_dict[ops[id].__name__]
            params['im'] = im
            if np.random.uniform(0, 1) < prob:
                im = ops[id](**params)
        im = np.asarray(im).astype('float32')
        if label is None:
            return (im, im_info)
        else:
            return (im, im_info, label)