kitti_transform.py

import torch
import pandas as pd
# from skimage import io, transform
import numpy as np
from torch.utils.data import Dataset, DataLoader
from torchvision import transforms, utils
from PIL import Image, ImageOps
import collections
try:
    import accimage
except ImportError:
    accimage = None
import random
import math
import numbers
import types
import scipy.ndimage as ndimage

import pdb


def _is_pil_image(img):
    if accimage is not None:
        return isinstance(img, (Image.Image, accimage.Image))
    else:
        return isinstance(img, Image.Image)


def _is_numpy_image(img):
    return isinstance(img, np.ndarray) and (img.ndim in {2, 3})

class RandomCrop(object):
    """Crop the given PIL.Image at a random location.
    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.
        padding (int or sequence, optional): Optional padding on each border
            of the image. Default is 0, i.e no padding. If a sequence of length
            4 is provided, it is used to pad left, top, right, bottom borders
            respectively.
    """

    def __init__(self, size, size_depth):
       
        self.size = size
        self.size_depth = size_depth


    def crop(self, img1, img2):
        w, h = img1.size
        tw, th = self.size

        i = random.randint(0, w - tw)
        # j = random.randint(h - th,h)

        img1 = img1.crop((i, h-th, i + tw, h))
        img2 = img2.crop((i, h-th, i + tw, h))

        return img1, img2

    def __call__(self, sample):
        """
        Args:
            img (PIL.Image): Image to be cropped.
        Returns:
            PIL.Image: Cropped image.
        """
        image, depth = sample['image'], sample['depth']
        
        w, h = image.size

        image,depth = self.crop(image,depth)

        # ow, oh = self.size_depth
        # depth = depth.resize((ow, oh))
        return {'image': image, 'depth': depth}


class RandomHorizontalFlip(object):

    def __call__(self, sample):
        image, depth = sample['image'], sample['depth']

        if not _is_pil_image(image):
            raise TypeError(
                'img should be PIL Image. Got {}'.format(type(img)))
        if not _is_pil_image(depth):
            raise TypeError(
                'img should be PIL Image. Got {}'.format(type(depth)))

        if random.random() < 0.5:
            image = image.transpose(Image.FLIP_LEFT_RIGHT)
            depth = depth.transpose(Image.FLIP_LEFT_RIGHT)
        #     print('imagesize:',image.size)
        # print('after flip:',image)
        # if random.random() < 0.5:
        #     image = np.flip(image,axis=0).copy()
        #     depth = np.flip(depth,axis=0).copy()

        return {'image': image, 'depth': depth}



class ToTensor(object):
    """Convert a ``PIL.Image`` or ``numpy.ndarray`` to tensor.
    Converts a PIL.Image or numpy.ndarray (H x W x C) in the range
    [0, 255] to a torch.FloatTensor of shape (C x H x W) in the range [0.0, 1.0].
    """

    def __call__(self, sample):
        image, depth = sample['image'], sample['depth']
        """
        Args:
            pic (PIL.Image or numpy.ndarray): Image to be converted to tensor.
        Returns:
            Tensor: Converted image.
        """
       
        image = self.to_tensor(image)
   
        # depth_tmp = np.asarray(depth)
        # depth = np.reshape(depth_tmp, depth_tmp.shape + (1,)) * 100
        depth = self.to_tensor(depth) 
        depth = torch.true_divide(depth, 256)
        
        # zeros = torch.zeros(1, 38, 152).float()
        # depth = torch.cat((zeros,depth),1)

       
        return {'image': image, 'depth': depth}
       

    def to_tensor(self, pic):
        if not(_is_pil_image(pic) or _is_numpy_image(pic)):
            raise TypeError(
                'pic should be PIL Image or ndarray. Got {}'.format(type(pic)))

        if isinstance(pic, np.ndarray):
            # handle numpy array
            img = torch.from_numpy(pic.transpose((2, 0, 1)))
            # backward compatibility
            return img.float()
        #    return img.float()
            # pdb.set_trace()

        if accimage is not None and isinstance(pic, accimage.Image):
            nppic = np.zeros(
                [pic.channels, pic.height, pic.width], dtype=np.float32)
            pic.copyto(nppic)
            return torch.from_numpy(nppic)

        # handle PIL Image
        if pic.mode == 'I':
            img = torch.from_numpy(np.array(pic, np.int32, copy=False))
        elif pic.mode == 'I;16':
            img = torch.from_numpy(np.array(pic, np.int16, copy=False))
        else:
            img = torch.ByteTensor(
                torch.ByteStorage.from_buffer(pic.tobytes()))
        # PIL image mode: 1, L, P, I, F, RGB, YCbCr, RGBA, CMYK
        if pic.mode == 'YCbCr':
            nchannel = 3
        elif pic.mode == 'I;16':
            nchannel = 1
        else:
            nchannel = len(pic.mode)
        img = img.view(pic.size[1], pic.size[0], nchannel)
        # put it from HWC to CHW format
        # yikes, this transpose takes 80% of the loading time/CPU
        img = img.transpose(0, 1).transpose(0, 2).contiguous()
        if isinstance(img, torch.ByteTensor):
            return img.float().div(255)
        else:
            return img


class Lighting(object):

    def __init__(self, alphastd, eigval, eigvec):
        self.alphastd = alphastd
        self.eigval = eigval
        self.eigvec = eigvec

    def __call__(self, sample):
        image, depth = sample['image'], sample['depth']
        if self.alphastd == 0:
            return image

        alpha = image.new().resize_(3).normal_(0, self.alphastd)
        rgb = self.eigvec.type_as(image).clone()\
            .mul(alpha.view(1, 3).expand(3, 3))\
            .mul(self.eigval.view(1, 3).expand(3, 3))\
            .sum(1).squeeze()

        image = image.add(rgb.view(3, 1, 1).expand_as(image))

        return {'image': image, 'depth': depth}


class Grayscale(object):

    def __call__(self, img):
        gs = img.clone()
        gs[0].mul_(0.299).add_(0.587, gs[1]).add_(0.114, gs[2])
        gs[1].copy_(gs[0])
        gs[2].copy_(gs[0])
        return gs


class Saturation(object):

    def __init__(self, var):
        self.var = var

    def __call__(self, img):
        gs = Grayscale()(img)
        alpha = random.uniform(-self.var, self.var)
        return img.lerp(gs, alpha)


class Brightness(object):

    def __init__(self, var):
        self.var = var

    def __call__(self, img):
        gs = img.new().resize_as_(img).zero_()
        alpha = random.uniform(-self.var, self.var)
        # print('brightness:',img,gs,alpha,img.lerp(gs, alpha))
        return img.lerp(gs, alpha)


class Contrast(object):

    def __init__(self, var):
        self.var = var

    def __call__(self, img):
        gs = Grayscale()(img)
        gs.fill_(gs.mean())
        alpha = random.uniform(-self.var, self.var)
        return img.lerp(gs, alpha)


class RandomOrder(object):
    """ Composes several transforms together in random order.
    """

    def __init__(self, transforms):
        self.transforms = transforms

    def __call__(self, sample):
        image, depth = sample['image'], sample['depth']

        if self.transforms is None:
            return {'image': image, 'depth': depth}
        order = torch.randperm(len(self.transforms))
        for i in order:
            image = self.transforms[i](image)

        return {'image': image, 'depth': depth}


class ColorJitter(RandomOrder):

    def __init__(self, brightness=0.4, contrast=0.4, saturation=0.4):
        self.transforms = []
        if brightness != 0:
            self.transforms.append(Brightness(brightness))
        if contrast != 0:
            self.transforms.append(Contrast(contrast))
        if saturation != 0:
            self.transforms.append(Saturation(saturation))


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

    def __init__(self, mean, std):
        self.mean = mean
        self.std = std

    def __call__(self, sample):
        """
        Args:
            tensor (Tensor): Tensor image of size (C, H, W) to be normalized.
        Returns:
            Tensor: Normalized image.
        """
        image, depth = sample['image'], sample['depth']

        image = self.normalize(image, self.mean, self.std)

        return {'image': image, 'depth': depth}

    def normalize(self, tensor, mean, std):
        """Normalize a tensor image with mean and standard deviation.
        See ``Normalize`` for more details.
        Args:
            tensor (Tensor): Tensor image of size (C, H, W) to be normalized.
            mean (sequence): Sequence of means for R, G, B channels respecitvely.
            std (sequence): Sequence of standard deviations for R, G, B channels
                respecitvely.
        Returns:
            Tensor: Normalized image.
        """

        # TODO: make efficient
        for t, m, s in zip(tensor, mean, std):
            # print('before normalize',t,m,s)
            t.sub_(m).div_(s)
            # print('after normalize',t,m,s)
        return tensor