How are image data handled in PyTorch?

1 year ago
Olivia Parker
1 minute

When working with image data in PyTorch, it is common to use the torchvision library, which offers convenient tools and functions for loading, processing, and transforming image data. Here are some commonly used methods for image data processing:

  1. Load image dataset.
from torchvision import datasets

# 加载训练集
train_dataset = datasets.ImageFolder('path/to/train_data', transform=transform)
# 加载测试集
test_dataset = datasets.ImageFolder('path/to/test_data', transform=transform)
  1. Image data preprocessing:
from torchvision import transforms

# 定义数据预处理步骤
transform = transforms.Compose([
    transforms.Resize(256),
    transforms.CenterCrop(224),
    transforms.ToTensor(),
    transforms.Normalize(mean=[0.485, 0.456, 0.406], std=[0.229, 0.224, 0.225])
])
  1. Load dataset into DataLoader.
from torch.utils.data import DataLoader

batch_size = 32

# 创建训练集和测试集的DataLoader
train_loader = DataLoader(train_dataset, batch_size=batch_size, shuffle=True)
test_loader = DataLoader(test_dataset, batch_size=batch_size, shuffle=False)
  1. Visualized image data:
import matplotlib.pyplot as plt

# 显示一张图像
image, label = train_dataset[0]
plt.imshow(image.permute(1, 2, 0))
plt.show()

With the above methods, it becomes easy to load, process, and visualize image data, providing convenience for model training and evaluation.

bannerAds