pytorch-image-models/timm/data/transforms.py

import torch
from torchvision import transforms
import torchvision.transforms.functional as F
from PIL import Image
import warnings
import math
import random
import numpy as np

from .constants import DEFAULT_CROP_PCT, IMAGENET_DEFAULT_MEAN, IMAGENET_DEFAULT_STD
from .random_erasing import RandomErasing
from .auto_augment import AutoAugment, auto_augment_policy


class ToNumpy:

    def __call__(self, pil_img):
        np_img = np.array(pil_img, dtype=np.uint8)
        if np_img.ndim < 3:
            np_img = np.expand_dims(np_img, axis=-1)
        np_img = np.rollaxis(np_img, 2)  # HWC to CHW
        return np_img


class ToTensor:

    def __init__(self, dtype=torch.float32):
        self.dtype = dtype

    def __call__(self, pil_img):
        np_img = np.array(pil_img, dtype=np.uint8)
        if np_img.ndim < 3:
            np_img = np.expand_dims(np_img, axis=-1)
        np_img = np.rollaxis(np_img, 2)  # HWC to CHW
        return torch.from_numpy(np_img).to(dtype=self.dtype)


_pil_interpolation_to_str = {
    Image.NEAREST: 'PIL.Image.NEAREST',
    Image.BILINEAR: 'PIL.Image.BILINEAR',
    Image.BICUBIC: 'PIL.Image.BICUBIC',
    Image.LANCZOS: 'PIL.Image.LANCZOS',
    Image.HAMMING: 'PIL.Image.HAMMING',
    Image.BOX: 'PIL.Image.BOX',
}


def _pil_interp(method):
    if method == 'bicubic':
        return Image.BICUBIC
    elif method == 'lanczos':
        return Image.LANCZOS
    elif method == 'hamming':
        return Image.HAMMING
    else:
        # default bilinear, do we want to allow nearest?
        return Image.BILINEAR


_RANDOM_INTERPOLATION = (Image.BILINEAR, Image.BICUBIC)


class RandomResizedCropAndInterpolation:
    """Crop the given PIL Image to random size and aspect ratio with random interpolation.

    A crop of random size (default: of 0.08 to 1.0) of the original size and a random
    aspect ratio (default: of 3/4 to 4/3) of the original aspect ratio is made. This crop
    is finally resized to given size.
    This is popularly used to train the Inception networks.

    Args:
        size: expected output size of each edge
        scale: range of size of the origin size cropped
        ratio: range of aspect ratio of the origin aspect ratio cropped
        interpolation: Default: PIL.Image.BILINEAR
    """

    def __init__(self, size, scale=(0.08, 1.0), ratio=(3. / 4., 4. / 3.),
                 interpolation='bilinear'):
        if isinstance(size, tuple):
            self.size = size
        else:
            self.size = (size, size)
        if (scale[0] > scale[1]) or (ratio[0] > ratio[1]):
            warnings.warn("range should be of kind (min, max)")

        if interpolation == 'random':
            self.interpolation = _RANDOM_INTERPOLATION
        else:
            self.interpolation = _pil_interp(interpolation)
        self.scale = scale
        self.ratio = ratio

    @staticmethod
    def get_params(img, scale, ratio):
        """Get parameters for ``crop`` for a random sized crop.

        Args:
            img (PIL Image): Image to be cropped.
            scale (tuple): range of size of the origin size cropped
            ratio (tuple): range of aspect ratio of the origin aspect ratio cropped

        Returns:
            tuple: params (i, j, h, w) to be passed to ``crop`` for a random
                sized crop.
        """
        area = img.size[0] * img.size[1]

        for attempt in range(10):
            target_area = random.uniform(*scale) * area
            log_ratio = (math.log(ratio[0]), math.log(ratio[1]))
            aspect_ratio = math.exp(random.uniform(*log_ratio))

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

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

        # Fallback to central crop
        in_ratio = img.size[0] / img.size[1]
        if in_ratio < min(ratio):
            w = img.size[0]
            h = int(round(w / min(ratio)))
        elif in_ratio > max(ratio):
            h = img.size[1]
            w = int(round(h * max(ratio)))
        else:  # whole image
            w = img.size[0]
            h = img.size[1]
        i = (img.size[1] - h) // 2
        j = (img.size[0] - w) // 2
        return i, j, h, w

    def __call__(self, img):
        """
        Args:
            img (PIL Image): Image to be cropped and resized.

        Returns:
            PIL Image: Randomly cropped and resized image.
        """
        i, j, h, w = self.get_params(img, self.scale, self.ratio)
        if isinstance(self.interpolation, (tuple, list)):
            interpolation = random.choice(self.interpolation)
        else:
            interpolation = self.interpolation
        return F.resized_crop(img, i, j, h, w, self.size, interpolation)

    def __repr__(self):
        if isinstance(self.interpolation, (tuple, list)):
            interpolate_str = ' '.join([_pil_interpolation_to_str[x] for x in self.interpolation])
        else:
            interpolate_str = _pil_interpolation_to_str[self.interpolation]
        format_string = self.__class__.__name__ + '(size={0}'.format(self.size)
        format_string += ', scale={0}'.format(tuple(round(s, 4) for s in self.scale))
        format_string += ', ratio={0}'.format(tuple(round(r, 4) for r in self.ratio))
        format_string += ', interpolation={0})'.format(interpolate_str)
        return format_string


def transforms_imagenet_train(
        img_size=224,
        scale=(0.08, 1.0),
        color_jitter=0.4,
        auto_augment=None,
        interpolation='random',
        random_erasing=0.4,
        random_erasing_mode='const',
        use_prefetcher=False,
        mean=IMAGENET_DEFAULT_MEAN,
        std=IMAGENET_DEFAULT_STD
):
    tfl = [
        RandomResizedCropAndInterpolation(
            img_size, scale=scale, interpolation=interpolation),
        transforms.RandomHorizontalFlip()
    ]
    if auto_augment:
        if isinstance(img_size, tuple):
            img_size_min = min(img_size)
        else:
            img_size_min = img_size
        aa_params = dict(
            translate_const=int(img_size_min * 0.45),
            img_mean=tuple([min(255, round(255 * x)) for x in mean]),
        )
        if interpolation and interpolation != 'random':
            aa_params['interpolation'] = _pil_interp(interpolation)
        aa_policy = auto_augment_policy(auto_augment, aa_params)
        tfl += [AutoAugment(aa_policy)]
    else:
        # color jitter is enabled when not using AA
        if isinstance(color_jitter, (list, tuple)):
            # color jitter should be a 3-tuple/list if spec brightness/contrast/saturation
            # or 4 if also augmenting hue
            assert len(color_jitter) in (3, 4)
        else:
            # if it's a scalar, duplicate for brightness, contrast, and saturation, no hue
            color_jitter = (float(color_jitter),) * 3
        tfl += [transforms.ColorJitter(*color_jitter)]

    if use_prefetcher:
        # prefetcher and collate will handle tensor conversion and norm
        tfl += [ToNumpy()]
    else:
        tfl += [
            transforms.ToTensor(),
            transforms.Normalize(
                mean=torch.tensor(mean),
                std=torch.tensor(std))
        ]
        if random_erasing > 0.:
            tfl.append(RandomErasing(random_erasing, mode=random_erasing_mode, device='cpu'))
    return transforms.Compose(tfl)


def transforms_imagenet_eval(
        img_size=224,
        crop_pct=None,
        interpolation='bilinear',
        use_prefetcher=False,
        mean=IMAGENET_DEFAULT_MEAN,
        std=IMAGENET_DEFAULT_STD):
    crop_pct = crop_pct or DEFAULT_CROP_PCT

    if isinstance(img_size, tuple):
        assert len(img_size) == 2
        if img_size[-1] == img_size[-2]:
            # fall-back to older behaviour so Resize scales to shortest edge if target is square
            scale_size = int(math.floor(img_size[0] / crop_pct))
        else:
            scale_size = tuple([int(x / crop_pct) for x in img_size])
    else:
        scale_size = int(math.floor(img_size / crop_pct))

    tfl = [
        transforms.Resize(scale_size, _pil_interp(interpolation)),
        transforms.CenterCrop(img_size),
    ]
    if use_prefetcher:
        # prefetcher and collate will handle tensor conversion and norm
        tfl += [ToNumpy()]
    else:
        tfl += [
            transforms.ToTensor(),
            transforms.Normalize(
                     mean=torch.tensor(mean),
                     std=torch.tensor(std))
        ]

    return transforms.Compose(tfl)