Unknown Error when using albumentations with CIFAR10 dataset(torchvision))

vision
1

I want to use albumentations library to perform some augmentations on predefined CIFAR10 dataset from torchvision. This is my code for writing the dataloaders

import torch
import torchvision
import torchvision.transforms as transforms

# Albumentations for augmentations
import albumentations as A
from albumentations.pytorch import ToTensorV2

train_transforms = A.Compose(
    [
        A.HorizontalFlip(p=0.5),
        A.ShiftScaleRotate(shift_limit=0.0625, scale_limit=0.1, rotate_limit=45, p=0.5),
        A.CoarseDropout(max_holes = 1, max_height=16, max_width=16, min_holes = 1, min_height=16, min_width=16,
                        fill_value=0.4734),
        A.Normalize(
            mean = (0.4914, 0.4822, 0.4465),
            std = (0.2470, 0.2435, 0.2616),
            p =1.0
        ),
        ToTensorV2()
    ],
    p=1.0
)


test_transforms = A.Compose(
    [
        A.Normalize(
            mean = (0.4914, 0.4822, 0.4465),
            std = (0.2470, 0.2435, 0.2616),
            p =1.0
        ),
        ToTensorV2()
    ]
)
torch.manual_seed(2)

use_cuda = torch.cuda.is_available()
device = torch.device("cuda" if use_cuda else "cpu")

class args():
    def __init__(self,device = 'cpu' ,use_cude = False) -> None:
        self.batch_size = 128
        self.device = device
        self.kwargs = {'num_workers': 1, 'pin_memory': True} if use_cuda else {}


# transform = transforms.Compose(
#     [transforms.ToTensor(),
#      transforms.Normalize((0.4914, 0.4822, 0.4465), (0.2470, 0.2435, 0.2616))])

trainset = torchvision.datasets.CIFAR10(root='./data', train=True,
                                        download=True, transform=train_transforms)
                                        
trainloader = torch.utils.data.DataLoader(trainset, batch_size=args().batch_size,
                                          shuffle=True, **args().kwargs)

testset = torchvision.datasets.CIFAR10(root='./data', train=False,
                                       download=True, transform=test_transforms)

testloader = torch.utils.data.DataLoader(testset, batch_size=args().batch_size,
                                         shuffle=False, **args().kwargs)

When I run the following code

tr = next(iter(trainloader))
print(len(tr))

It gives the following error -
AssertionError: force_apply must have bool or int type

Currenty running this on colab

Can I not use albumentations for the predefined datasets on Torchvision ? Will I have to write a custom dataset for CIfar10 so that I can use albumentations for augmentations?

1 Like
2

Never Mind. This is how I fixed it if anyone faces the same problem.

!pip install albumentations==0.4.6

import cv2

import torchvision

import torch

import torchvision.transforms as transforms

# Albumentations for augmentations

import albumentations as A

from albumentations.pytorch import ToTensorV2

cv2.setNumThreads(0)

cv2.ocl.setUseOpenCL(False)

class Cifar10SearchDataset(torchvision.datasets.CIFAR10):

    def __init__(self, root="~/data/cifar10", train=True, download=True, transform=None):

        super().__init__(root=root, train=train, download=download, transform=transform)

    def __getitem__(self, index):

        image, label = self.data[index], self.targets[index]

        if self.transform is not None:

            transformed = self.transform(image=image)

            image = transformed["image"]

        return image, label

train_transforms = A.Compose(

    [

        A.HorizontalFlip(p=0.5),

        A.ShiftScaleRotate(shift_limit=0.0625, scale_limit=0.1, rotate_limit=45, p=0.5),

        A.CoarseDropout(max_holes = 1, max_height=16, max_width=16, min_holes = 1, min_height=16, min_width=16,

                        fill_value=0.4734),

        A.Normalize(

            mean = (0.4914, 0.4822, 0.4465),

            std = (0.2470, 0.2435, 0.2616),

            p =1.0

        ),

        ToTensorV2()

    ],

    p=1.0

)

class args():

    def __init__(self,device = 'cpu' ,use_cuda = False) -> None:

        self.batch_size = 128

        self.device = device

        self.use_cuda = use_cuda

        self.kwargs = {'num_workers': 1, 'pin_memory': True} if self.use_cuda else {}

trainset = Cifar10SearchDataset(root='./data', train=True,

                                        download=True, transform=train_transforms)

                                        

trainloader = torch.utils.data.DataLoader(trainset, batch_size=args().batch_size,

                                          shuffle=True, **args().kwargs)
1 Like