詳解Python實現圖像分割增強的兩種方法

import random
import numpy as np
from PIL import Image, ImageOps, ImageFilter
from skimage.filters import gaussian
import torch
import math
import numbers
import random

class RandomVerticalFlip(object):
    def __call__(self, img):
        if random.random() < 0.5:
            return img.transpose(Image.FLIP_TOP_BOTTOM)
        return img

class DeNormalize(object):
    def __init__(self, mean, std):
        self.mean = mean
        self.std = std

    def __call__(self, tensor):
        for t, m, s in zip(tensor, self.mean, self.std):
            t.mul_(s).add_(m)
        return tensor

class MaskToTensor(object):
    def __call__(self, img):
        return torch.from_numpy(np.array(img, dtype=np.int32)).long()

class FreeScale(object):
    def __init__(self, size, interpolation=Image.BILINEAR):
        self.size = tuple(reversed(size))  # size: (h, w)
        self.interpolation = interpolation

    def __call__(self, img):
        return img.resize(self.size, self.interpolation)

class FlipChannels(object):
    def __call__(self, img):
        img = np.array(img)[:, :, ::-1]
        return Image.fromarray(img.astype(np.uint8))

class RandomGaussianBlur(object):
    def __call__(self, img):
        sigma = 0.15 + random.random() * 1.15
        blurred_img = gaussian(np.array(img), sigma=sigma, multichannel=True)
        blurred_img *= 255
        return Image.fromarray(blurred_img.astype(np.uint8))
# 組合
class Compose(object):
    def __init__(self, transforms):
        self.transforms = transforms

    def __call__(self, img, mask):
        assert img.size == mask.size
        for t in self.transforms:
            img, mask = t(img, mask)
        return img, mask
# 隨機裁剪
class RandomCrop(object):
    def __init__(self, size, padding=0):
        if isinstance(size, numbers.Number):
            self.size = (int(size), int(size))
        else:
            self.size = size
        self.padding = padding

    def __call__(self, img, mask):
        if self.padding > 0:
            img = ImageOps.expand(img, border=self.padding, fill=0)
            mask = ImageOps.expand(mask, border=self.padding, fill=0)

        assert img.size == mask.size
        w, h = img.size
        th, tw = self.size
        if w == tw and h == th:
            return img, mask
        if w < tw or h < th:
            return img.resize((tw, th), Image.BILINEAR), mask.resize((tw, th), Image.NEAREST)

        x1 = random.randint(0, w - tw)
        y1 = random.randint(0, h - th)
        return img.crop((x1, y1, x1 + tw, y1 + th)), mask.crop((x1, y1, x1 + tw, y1 + th))

#  中心裁剪
class CenterCrop(object):
    def __init__(self, size):
        if isinstance(size, numbers.Number):
            self.size = (int(size), int(size))
        else:
            self.size = size

    def __call__(self, img, mask):
        assert img.size == mask.size
        w, h = img.size
        th, tw = self.size
        x1 = int(round((w - tw) / 2.))
        y1 = int(round((h - th) / 2.))
        return img.crop((x1, y1, x1 + tw, y1 + th)), mask.crop((x1, y1, x1 + tw, y1 + th))


class RandomHorizontallyFlip(object):
    def __call__(self, img, mask):
        if random.random() < 0.5:
            return img.transpose(Image.FLIP_LEFT_RIGHT), mask.transpose(Image.FLIP_LEFT_RIGHT)
        return img, mask

class Scale(object):
    def __init__(self, size):
        self.size = size

    def __call__(self, img, mask):
        assert img.size == mask.size
        w, h = img.size
        if (w >= h and w == self.size) or (h >= w and h == self.size):
            return img, mask
        if w > h:
            ow = self.size
            oh = int(self.size * h / w)
            return img.resize((ow, oh), Image.BILINEAR), mask.resize((ow, oh), Image.NEAREST)
        else:
            oh = self.size
            ow = int(self.size * w / h)
            return img.resize((ow, oh), Image.BILINEAR), mask.resize((ow, oh), Image.NEAREST)

class RandomSizedCrop(object):
    def __init__(self, size):
        self.size = size

    def __call__(self, img, mask):
        assert img.size == mask.size
        for attempt in range(10):
            area = img.size[0] * img.size[1]
            target_area = random.uniform(0.45, 1.0) * area
            aspect_ratio = random.uniform(0.5, 2)

            w = int(round(math.sqrt(target_area * aspect_ratio)))
            h = int(round(math.sqrt(target_area / aspect_ratio)))

            if random.random() < 0.5:
                w, h = h, w

            if w <= img.size[0] and h <= img.size[1]:
                x1 = random.randint(0, img.size[0] - w)
                y1 = random.randint(0, img.size[1] - h)

                img = img.crop((x1, y1, x1 + w, y1 + h))
                mask = mask.crop((x1, y1, x1 + w, y1 + h))
                assert (img.size == (w, h))

                return img.resize((self.size, self.size), Image.BILINEAR), mask.resize((self.size, self.size),
                                                                                       Image.NEAREST)

        # Fallback
        scale = Scale(self.size)
        crop = CenterCrop(self.size)
        return crop(*scale(img, mask))

class RandomRotate(object):
    def __init__(self, degree):
        self.degree = degree

    def __call__(self, img, mask):
        rotate_degree = random.random() * 2 * self.degree - self.degree
        return img.rotate(rotate_degree, Image.BILINEAR), mask.rotate(rotate_degree, Image.NEAREST)

class RandomSized(object):
    def __init__(self, size):
        self.size = size
        self.scale = Scale(self.size)
        self.crop = RandomCrop(self.size)

    def __call__(self, img, mask):
        assert img.size == mask.size

        w = int(random.uniform(0.5, 2) * img.size[0])
        h = int(random.uniform(0.5, 2) * img.size[1])

        img, mask = img.resize((w, h), Image.BILINEAR), mask.resize((w, h), Image.NEAREST)

        return self.crop(*self.scale(img, mask))

class SlidingCropOld(object):
    def __init__(self, crop_size, stride_rate, ignore_label):
        self.crop_size = crop_size
        self.stride_rate = stride_rate
        self.ignore_label = ignore_label

    def _pad(self, img, mask):
        h, w = img.shape[: 2]
        pad_h = max(self.crop_size - h, 0)
        pad_w = max(self.crop_size - w, 0)
        img = np.pad(img, ((0, pad_h), (0, pad_w), (0, 0)), 'constant')
        mask = np.pad(mask, ((0, pad_h), (0, pad_w)), 'constant', constant_values=self.ignore_label)
        return img, mask

    def __call__(self, img, mask):
        assert img.size == mask.size

        w, h = img.size
        long_size = max(h, w)

        img = np.array(img)
        mask = np.array(mask)

        if long_size > self.crop_size:
            stride = int(math.ceil(self.crop_size * self.stride_rate))
            h_step_num = int(math.ceil((h - self.crop_size) / float(stride))) + 1
            w_step_num = int(math.ceil((w - self.crop_size) / float(stride))) + 1
            img_sublist, mask_sublist = [], []
            for yy in range(h_step_num):
                for xx in range(w_step_num):
                    sy, sx = yy * stride, xx * stride
                    ey, ex = sy + self.crop_size, sx + self.crop_size
                    img_sub = img[sy: ey, sx: ex, :]
                    mask_sub = mask[sy: ey, sx: ex]
                    img_sub, mask_sub = self._pad(img_sub, mask_sub)
                    img_sublist.append(Image.fromarray(img_sub.astype(np.uint8)).convert('RGB'))
                    mask_sublist.append(Image.fromarray(mask_sub.astype(np.uint8)).convert('P'))
            return img_sublist, mask_sublist
        else:
            img, mask = self._pad(img, mask)
            img = Image.fromarray(img.astype(np.uint8)).convert('RGB')
            mask = Image.fromarray(mask.astype(np.uint8)).convert('P')
            return img, mask


class SlidingCrop(object):
    def __init__(self, crop_size, stride_rate, ignore_label):
        self.crop_size = crop_size
        self.stride_rate = stride_rate
        self.ignore_label = ignore_label

    def _pad(self, img, mask):
        h, w = img.shape[: 2]
        pad_h = max(self.crop_size - h, 0)
        pad_w = max(self.crop_size - w, 0)
        img = np.pad(img, ((0, pad_h), (0, pad_w), (0, 0)), 'constant')
        mask = np.pad(mask, ((0, pad_h), (0, pad_w)), 'constant', constant_values=self.ignore_label)
        return img, mask, h, w

    def __call__(self, img, mask):
        assert img.size == mask.size

        w, h = img.size
        long_size = max(h, w)

        img = np.array(img)
        mask = np.array(mask)

        if long_size > self.crop_size:
            stride = int(math.ceil(self.crop_size * self.stride_rate))
            h_step_num = int(math.ceil((h - self.crop_size) / float(stride))) + 1
            w_step_num = int(math.ceil((w - self.crop_size) / float(stride))) + 1
            img_slices, mask_slices, slices_info = [], [], []
            for yy in range(h_step_num):
                for xx in range(w_step_num):
                    sy, sx = yy * stride, xx * stride
                    ey, ex = sy + self.crop_size, sx + self.crop_size
                    img_sub = img[sy: ey, sx: ex, :]
                    mask_sub = mask[sy: ey, sx: ex]
                    img_sub, mask_sub, sub_h, sub_w = self._pad(img_sub, mask_sub)
                    img_slices.append(Image.fromarray(img_sub.astype(np.uint8)).convert('RGB'))
                    mask_slices.append(Image.fromarray(mask_sub.astype(np.uint8)).convert('P'))
                    slices_info.append([sy, ey, sx, ex, sub_h, sub_w])
            return img_slices, mask_slices, slices_info
        else:
            img, mask, sub_h, sub_w = self._pad(img, mask)
            img = Image.fromarray(img.astype(np.uint8)).convert('RGB')
            mask = Image.fromarray(mask.astype(np.uint8)).convert('P')
            return [img], [mask], [[0, sub_h, 0, sub_w, sub_h, sub_w]]

import numpy as np
import random

import torch
from torchvision import transforms as T
from torchvision.transforms import functional as F


def pad_if_smaller(img, size, fill=0):
    # 如果圖像最小邊長小於給定size，則用數值fill進行padding
    min_size = min(img.size)
    if min_size < size:
        ow, oh = img.size
        padh = size - oh if oh < size else 0
        padw = size - ow if ow < size else 0
        img = F.pad(img, (0, 0, padw, padh), fill=fill)
    return img


class Compose(object):
    def __init__(self, transforms):
        self.transforms = transforms

    def __call__(self, image, target):
        for t in self.transforms:
            image, target = t(image, target)
        return image, target


class RandomResize(object):
    def __init__(self, min_size, max_size=None):
        self.min_size = min_size
        if max_size is None:
            max_size = min_size
        self.max_size = max_size

    def __call__(self, image, target):
        size = random.randint(self.min_size, self.max_size)
        # 這裡size傳入的是int類型，所以是將圖像的最小邊長縮放到size大小
        image = F.resize(image, size)
        # 這裡的interpolation註意下，在torchvision(0.9.0)以後才有InterpolationMode.NEAREST
        # 如果是之前的版本需要使用PIL.Image.NEAREST
        target = F.resize(target, size, interpolation=T.InterpolationMode.NEAREST)
        return image, target


class RandomHorizontalFlip(object):
    def __init__(self, flip_prob):
        self.flip_prob = flip_prob

    def __call__(self, image, target):
        if random.random() < self.flip_prob:
            image = F.hflip(image)
            target = F.hflip(target)
        return image, target


class RandomCrop(object):
    def __init__(self, size):
        self.size = size

    def __call__(self, image, target):
        image = pad_if_smaller(image, self.size)
        target = pad_if_smaller(target, self.size, fill=255)
        crop_params = T.RandomCrop.get_params(image, (self.size, self.size))
        image = F.crop(image, *crop_params)
        target = F.crop(target, *crop_params)
        return image, target


class CenterCrop(object):
    def __init__(self, size):
        self.size = size

    def __call__(self, image, target):
        image = F.center_crop(image, self.size)
        target = F.center_crop(target, self.size)
        return image, target


class ToTensor(object):
    def __call__(self, image, target):
        image = F.to_tensor(image)
        target = torch.as_tensor(np.array(target), dtype=torch.int64)
        return image, target


class Normalize(object):
    def __init__(self, mean, std):
        self.mean = mean
        self.std = std

    def __call__(self, image, target):
        image = F.normalize(image, mean=self.mean, std=self.std)
        return image, target