補充:transform.invert 預處理逆操作
from PIL import Image
from torchvision import transforms
import torch
import numpy as np
def transform_invert(img_, transform_train):
"""
将data 進行反transfrom操作
:param img_: tensor
:param transform_train: torchvision.transforms
:return: PIL image
"""
if 'Normalize' in str(transform_train):
# 分析transforms裡的Normalize
norm_transform = list(filter(lambda x: isinstance(x, transforms.Normalize), transform_train.transforms))
mean = torch.tensor(norm_transform[0].mean, dtype=img_.dtype, device=img_.device)
std = torch.tensor(norm_transform[0].std, dtype=img_.dtype, device=img_.device)
img_.mul_(std[:, None, None]).add_(mean[:, None, None]) # 廣播三個次元 乘标準差 加均值
img_ = img_.transpose(0, 2).transpose(0, 1) # C*H*W --> H*W*C
# 如果有ToTensor,那麼之前數值就會被壓縮至0-1之間。現在需要反變換回來,也就是乘255
if 'ToTensor' in str(transform_train):
img_ = np.array(img_) * 255
# 先将np的元素轉換為uint8資料類型,然後轉換為PIL.Image類型
if img_.shape[2] == 3: # 若通道數為3 需要轉為RGB類型
img_ = Image.fromarray(img_.astype('uint8')).convert('RGB')
elif img_.shape[2] == 1: # 若通道數為1 需要壓縮張量的次元至2D
img_ = Image.fromarray(img_.astype('uint8').squeeze())
else:
raise Exception("Invalid img shape, expected 1 or 3 in axis 2, but got {}!".format(img_.shape[2]))
return img_
if __name__ == '__main__':
img = Image.open(r"./test.jpg").convert('RGB')
img_transform = transforms.Compose([transforms.ToTensor()])
img_tensor = img_transform(img)
# img_tensor.unsqueeze_(dim=0) # C*H*W to B*C*H*W
print(img_tensor)
print(img_tensor.shape)
img = transform_invert(img_tensor, img_transform) # input: shape=[c h w]
img.show()
随機擦除: RandomErasing
功能:對圖像進行随機遮擋
主要參數說明:
- p:機率值,執行該操作的機率
- scale:遮擋區域的面積
- ratio:遮擋區域長寬比(a:長比寬,b:寬比長)
- value:設定遮擋區域的像素值,(R, G, B) or (Gray)
from PIL import Image
from torchvision import transforms
from utils import transform_invert
if __name__ == '__main__':
# 1.讀取圖像
img = Image.open(r"./cat.png").convert('RGB')
# 2.确定預處理方式
img_transform = transforms.Compose([## transforms.Resize((300,300)), # 重置大小為300*300
transforms.ToTensor() , # 轉Tensor型變量
transforms.RandomErasing(p=1,scale=(0.01,0.35),ratio=(0.5,1),value=(0,0,255)), # 仿射變換
])
# 3.進行預處理
img_tensor = img_transform(img)
# 4.逆Transform變換
img = transform_invert(img_tensor, img_transform) # input: shape=[c h w]
# 5.進行預處理效果展示
img.show()