torch tensor左右翻转

完整例子：

import torch
import torchvision.transforms as transforms
import torchvision.transforms.functional as F
from PIL import Image

# Step 1: 读取一张图片
image_path = r"C:\Users\Administrator\Pictures\pinije\3.jpg"  # 替换为你的图片路径
img = Image.open(image_path)

# Step 2: 转为Tensor
transform_to_tensor = transforms.ToTensor()
img_tensor = transform_to_tensor(img)

batch_size = 4
batch_img_tensor = img_tensor.unsqueeze(0).repeat(batch_size, 1, 1, 1)

# Step 4: 翻转
flipped_batch_tensor = torch.flip(batch_img_tensor, [-1])

# Step 5: 转回图片并保存（仅保存第一个批量数据）
transform_to_pil = transforms.ToPILImage()
flipped_img = transform_to_pil(flipped_batch_tensor[0])

# 保存图片
flipped_img.save("flipped_image.jpg")

import torch
import torchvision.transforms as transforms

# 假设你的图像张量是 img_tensor
# img_tensor 的尺寸是 (1, 3, 256, 448)

# 使用 transforms.RandomHorizontalFlip()
transform = transforms.RandomHorizontalFlip(p=1)  # p=1 表示总是翻转
flipped_tensor = transform(img_tensor)

print(flipped_tensor.shape)

import torch
import torchvision.transforms.functional as F

# 假设你的图像张量是 img_tensor
# img_tensor 的尺寸是 (1, 3, 256, 448)

# 使用 F.hflip
flipped_tensor = F.hflip(img_tensor)

print(flipped_tensor.shape)

import torch
import torchvision.transforms as transforms
import torchvision.transforms.functional as F

# 创建一个示例图像张量，尺寸为 (1, 3, 256, 448)
img_tensor = torch.randn(1, 3, 256, 448)

# 方法1：使用 transforms.RandomHorizontalFlip
transform = transforms.RandomHorizontalFlip(p=1)  # p=1 表示总是翻转
flipped_tensor1 = transform(img_tensor)
print(f"使用 transforms.RandomHorizontalFlip 翻转后的张量尺寸: {flipped_tensor1.shape}")

# 方法2：使用 F.hflip
flipped_tensor2 = F.hflip(img_tensor)
print(f"使用 F.hflip 翻转后的张量尺寸: {flipped_tensor2.shape}")