Image data Modeling example

Prepare dataset

  • Using cifr-2 as an example
  • Import necessary pakcage
    1
    2
    3
    4
    5
    import torch 
    from torch import nn
    from torch.utils.data import Dataset,DataLoader
    from torchvision import transforms as T
    from torchvision import datasets 
    1
    2
    3
    4
    5
    6
    7
    8
    9
    10
    11
    12
    13
    14
    15
    16
    17
    # define helper function
    transform_img = T.Compose(
        [T.ToTensor()])

    def transform_label(x):
        return torch.tensor([x]).float()

    # load dataset using Imagefolder
    ds_train = datasets.ImageFolder("./eat_pytorch_datasets/cifar2/train/",
                transform = transform_img,target_transform = transform_label)
    ds_val = datasets.ImageFolder("./eat_pytorch_datasets/cifar2/test/",
                transform = transform_img,target_transform = transform_label)
    print(ds_train.class_to_idx)    

    # convert into data loader
    dl_train = DataLoader(ds_train,batch_size = 50,shuffle = True)
    dl_val = DataLoader(ds_val,batch_size = 50,shuffle = False)
  • The deault order of pytorch img is : (Batch,Channel,Width,Height)

Define the model

  • 是我喜欢的模型定义方式，直接背下来
    1
    2
    3
    4
    5
    6
    7
    8
    9
    10
    11
    12
    13
    14
    15
    16
    17
    18
    19
    20
    21
    22
    23
    24
    25
    26
    27
    28
    29
    class net(nn.Module):
        def __init__(net, self):
        # Define the architecture
            self.conv1 = nn.Conv2d(in_channels=3,out_channels=32,kernel_size = 3)
            self.pool = nn.MaxPool2d(kernel_size = 2,stride = 2)
            self.conv2 = nn.Conv2d(in_channels=32,out_channels=64,kernel_size = 5)
            self.dropout = nn.Dropout2d(p = 0.1)
            self.adaptive_pool = nn.AdaptiveMaxPool2d((1,1))
            self.flatten = nn.Flatten()
            self.linear1 = nn.Linear(64,32)
            self.relu = nn.ReLU()
            self.linear2 = nn.Linear(32,1)

        # Do the forward propagation
        def forward(self,x):
            x = self.conv1(x)
            x = self.pool(x)
            x = self.conv2(x)
            x = self.pool(x)
            x = self.dropout(x)
            x = self.adaptive_pool(x)
            x = self.flatten(x)
            x = self.linear1(x)
            x = self.relu(x)
            x = self.linear2(x)
            return x 
            
    net = Net()
    print(net)

Training the model

  • 不是我喜欢的style，直接不管

Something else

  • Saving the model, using the model, and eval the model follows the same procedure of Day01. omit here.