About transform in CNN

[GAuravY19]

train_data = datasets.FashionMNIST(
    root='data', # where to download the data  
    train = True, # do we want the training dataset ? 
    download = True, # do we want to download Yes/No
    transform = torchvision.transforms.ToTensor(), # how do we want to transform the data ?
    target_transform = None # How do we want to transform the labels/targets ? 
)


test_data = datasets.FashionMNIST(
    root = 'data',
    train = False,
    download = True,
    transform = ToTensor(),
    target_transform = None
)

My question :- What is the difference between torchvision.transforms.ToTensor() and transform = ToTensor()

[LuluW8071]

Both performs same function.

The only difference lies in how you import

import torchvision.transforms as transforms
import torch

# Using torchvision.transforms.ToTensor()
transform_1 = transforms.ToTensor()
image_tensor_1 = transform_1(pil_image)  

# Using transform = ToTensor()
from torchvision.transforms import ToTensor

transform_2 = ToTensor()
image_tensor_2 = transform_2(pil_image)

[GAuravY19]

Thanks for replying.....