[pytorch learning notes] transforms

Transforms

all torchvision Data sets have two parameters transform and target_transform, The former enhances the image , The latter deals with labels （ Label smoothing ）.Fashionmnist yes PIL Image format , The label is an integer . For training , It is necessary to convert the features into normalized tensor, And label one-hot Coding is convenient softmax. have access to ToTensor and Lambda.

import torch
from torchvision import datasets
from torchvision.transforms import ToTensor, Lambda

ds = datasets.FashionMNIST(
    root="data",
    train=True,
    download=True,
    transform=ToTensor(),
    target_transform=Lambda(lambda y: torch.zeros(10, dtype=torch.float).scatter_(0, torch.tensor(y), value=1))
)

ToTensor()

ToTensor() take PIL Image or ndarray Convert to FloatTensor, And normalize the pixels to [0,1]

Transforms pytorchAPI