pre_process.py

import numpy as np
from torchvision import transforms
import os
from PIL import Image, ImageOps
import numbers
import torch

class ResizeImage():
    def __init__(self, size):
      if isinstance(size, int):
        self.size = (int(size), int(size))
      else:
        self.size = size
    def __call__(self, img):
      th, tw = self.size
      return img.resize((th, tw))

class RandomSizedCrop(object):
    """Crop the given PIL.Image to random size and aspect ratio.
    A crop of random size of (0.08 to 1.0) of the original size and a random
    aspect ratio 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: size of the smaller edge
        interpolation: Default: PIL.Image.BILINEAR
    """

    def __init__(self, size, interpolation=Image.BILINEAR):
        self.size = size
        self.interpolation = interpolation

    def __call__(self, img):
        h_off = random.randint(0, img.shape[1]-self.size)
        w_off = random.randint(0, img.shape[2]-self.size)
        img = img[:, h_off:h_off+self.size, w_off:w_off+self.size]
        return img


class Normalize(object):
    """Normalize an tensor image with mean and standard deviation.
    Given mean: (R, G, B),
    will normalize each channel of the torch.*Tensor, i.e.
    channel = channel - mean
    Args:
        mean (sequence): Sequence of means for R, G, B channels respecitvely.
    """

    def __init__(self, mean=None, meanfile=None):
        if mean:
            self.mean = mean
        else:
            arr = np.load(meanfile)
            self.mean = torch.from_numpy(arr.astype('float32')/255.0)[[2,1,0],:,:]

    def __call__(self, tensor):
        """
        Args:
            tensor (Tensor): Tensor image of size (C, H, W) to be normalized.
        Returns:
            Tensor: Normalized image.
        """
        # TODO: make efficient
        for t, m in zip(tensor, self.mean):
            t.sub_(m)
        return tensor



class PlaceCrop(object):
    """Crops the given PIL.Image at the particular index.
    Args:
        size (sequence or int): Desired output size of the crop. If size is an
            int instead of sequence like (w, h), a square crop (size, size) is
            made.
    """

    def __init__(self, size, start_x, start_y):
        if isinstance(size, int):
            self.size = (int(size), int(size))
        else:
            self.size = size
        self.start_x = start_x
        self.start_y = start_y

    def __call__(self, img):
        """
        Args:
            img (PIL.Image): Image to be cropped.
        Returns:
            PIL.Image: Cropped image.
        """
        th, tw = self.size
        return img.crop((self.start_x, self.start_y, self.start_x + tw, self.start_y + th))


class ForceFlip(object):
    """Horizontally flip the given PIL.Image randomly with a probability of 0.5."""

    def __call__(self, img):
        """
        Args:
            img (PIL.Image): Image to be flipped.
        Returns:
            PIL.Image: Randomly flipped image.
        """
        return img.transpose(Image.FLIP_LEFT_RIGHT)

class CenterCrop(object):
    """Crops the given PIL.Image at the center.
    Args:
        size (sequence or int): Desired output size of the crop. If size is an
            int instead of sequence like (h, w), a square crop (size, size) is
            made.
    """

    def __init__(self, size):
        if isinstance(size, numbers.Number):
            self.size = (int(size), int(size))
        else:
            self.size = size

    def __call__(self, img):
        """
        Args:
            img (PIL.Image): Image to be cropped.
        Returns:
            PIL.Image: Cropped image.
        """
        w, h = (img.shape[1], img.shape[2])
        th, tw = self.size
        w_off = int((w - tw) / 2.)
        h_off = int((h - th) / 2.)
        img = img[:, h_off:h_off+th, w_off:w_off+tw]
        return img


def image_train(resize_size=256, crop_size=224, alexnet=False):
  if not alexnet:
    normalize = transforms.Normalize(mean=[0.485, 0.456, 0.406],
                                   std=[0.229, 0.224, 0.225])
  else:
    normalize = Normalize(meanfile='./ilsvrc_2012_mean.npy')
  return  transforms.Compose([
        ResizeImage(resize_size),
        transforms.RandomResizedCrop(crop_size),
        transforms.RandomHorizontalFlip(),
        transforms.ToTensor(),
        normalize
    ])

def image_test(resize_size=256, crop_size=224, alexnet=False):
  if not alexnet:
    normalize = transforms.Normalize(mean=[0.485, 0.456, 0.406],
                                   std=[0.229, 0.224, 0.225])
  else:
    normalize = Normalize(meanfile='./ilsvrc_2012_mean.npy')
  start_first = 0
  start_center = (resize_size - crop_size - 1) / 2
  start_last = resize_size - crop_size - 1
 
  return transforms.Compose([
    ResizeImage(resize_size),
    PlaceCrop(crop_size, start_center, start_center),
    transforms.ToTensor(),
    normalize
  ])

def image_test_10crop(resize_size=256, crop_size=224, alexnet=False):
    if not alexnet:
        normalize = transforms.Normalize(mean=[0.485, 0.456, 0.406],
                                   std=[0.229, 0.224, 0.225])
    else:
        normalize = Normalize(meanfile='./ilsvrc_2012_mean.npy')
    start_first = 0
    start_center = (resize_size - crop_size - 1) / 2
    start_last = resize_size - crop_size - 1
    data_transforms = [
        transforms.Compose([
        ResizeImage(resize_size),ForceFlip(),
        PlaceCrop(crop_size, start_first, start_first),
        transforms.ToTensor(),
        normalize
        ]),
        transforms.Compose([
        ResizeImage(resize_size),ForceFlip(),
        PlaceCrop(crop_size, start_last, start_last),
        transforms.ToTensor(),
        normalize
        ]),
        transforms.Compose([
        ResizeImage(resize_size),ForceFlip(),
        PlaceCrop(crop_size, start_last, start_first),
        transforms.ToTensor(),
        normalize
        ]),
        transforms.Compose([
        ResizeImage(resize_size),ForceFlip(),
        PlaceCrop(crop_size, start_first, start_last),
        transforms.ToTensor(),
        normalize
        ]),
        transforms.Compose([
        ResizeImage(resize_size),ForceFlip(),
        PlaceCrop(crop_size, start_center, start_center),
        transforms.ToTensor(),
        normalize
        ]),
        transforms.Compose([
        ResizeImage(resize_size),
        PlaceCrop(crop_size, start_first, start_first),
        transforms.ToTensor(),
        normalize
        ]),
        transforms.Compose([
        ResizeImage(resize_size),
        PlaceCrop(crop_size, start_last, start_last),
        transforms.ToTensor(),
        normalize
        ]),
        transforms.Compose([
        ResizeImage(resize_size),
        PlaceCrop(crop_size, start_last, start_first),
        transforms.ToTensor(),
        normalize
        ]),
        transforms.Compose([
        ResizeImage(resize_size),
        PlaceCrop(crop_size, start_first, start_last),
        transforms.ToTensor(),
        normalize
        ]),
        transforms.Compose([
        ResizeImage(resize_size),
        PlaceCrop(crop_size, start_center, start_center),
        transforms.ToTensor(),
        normalize
        ])
    ]
    return data_transforms

def inv_preprocess(imgs, num_images=1):
    """Inverse preprocessing of the batch of images.
       
    Args:
      imgs: batch of input images.
      num_images: number of images to apply the inverse transformations on.
      img_mean: vector of mean colour values.
      numpy_transform: whether change RGB to BGR during img_transform.
  
    Returns:
      The batch of the size num_images with the same spatial dimensions as the input.
    """
    def norm_ip(img, min, max):
        img.clamp_(min=min, max=max)
        img.add_(-min).div_(max - min + 1e-5)
    norm_ip(imgs, float(imgs.min()), float(imgs.max()))
    return imgs