实验算法设计

Unet

• 可以用双线性差值替换，效果差不多，参数更少。
  

from typing import Dict
import torch
import torch.nn as nn
import torch.nn.functional as F 

class DoubleConv(nn.Sequential):
    def __init__(self, in_channels, out_channels, mid_channels= None):
        if mid_channels is None:
            mid_channels = out_channels
        
        super(DoubleConv, self).__init__(
            nn.Conv2d(in_channels, mid_channels, kernel_size=3,padding=1,bias=False),
            nn.BatchNorm2d(mid_channels),
            nn.ReLU(inplace=True),
            nn.Conv2d(mid_channels, out_channels, kernel_size=3,padding=1,bias=False),
            nn.BatchNorm2d(out_channels),
            nn.ReLU(inplace=True)
        )

class Down(nn.Module):
    def __init__(self, in_channels, out_channels):
        super(Down,self).__init__(
            nn.MaxPool2d(2,stride=2),
            DoubleConv(in_channels, out_channels)
        )

class Up(nn.Module):
    def __init__(self, in_channels, out_channels, bilinear=True):
        super(Up, self).__init__()

        if bilinear:
            self.up = nn.Upsample(scale_factor=2, mode='bilinear', align_corners=True)
            self.conv = DoubleConv(in_channels, out_channels, in_channels // 2)
        else:
            self.up = nn.ConvTranspose2d(in_channels , in_channels // 2, kernel_size=2, stride=2)
            self.conv = DoubleConv(in_channels, out_channels)

    def forward(self, x1: torch.Tensor, x2: torch.Tensor)-> torch.Tensor:
        x1 = self.up(x1)
        diff_x = x2.size()[2] - x1.size()[2]
        diff_y = x2.size()[3] - x1.size()[3]

        x1 = F.pad(x1, [diff_x // 2, diff_x - diff_x // 2,
                        diff_y // 2, diff_y - diff_y // 2])
        x = torch.cat([x2, x1], dim=1)
        x = self.conv(x)
        return x

 class OutConv(nn.Sequential):
    def __init__(self, in_channels, num_classes):
        super(OutConv, self).__init__(
            nn.Conv2d(in_channels, num_classes, kernel_size=1),
        )

 class Unet(nn.Module):
    def __init__(self,
                 in_channels: int = 1, 
                 num_classes: int = 2, 
                 bilinear: bool = True,
                 base_c: int = 64
                 ):
        super(Unet, self).__init__()
        self.in_channels = in_channels
        self.num_classes = num_classes
        self.bilinear = bilinear
        self.in_conv = DoubleConv(in_channels, base_c)
        self.down1 = Down(base_c, base_c * 2)
        self.down2 = Down(base_c * 2, base_c * 4)
        self.down3 = Down(base_c * 4, base_c * 8)
        factor = 2 if bilinear else 1
        self.down4 = Down(base_c * 8, base_c * 16 // factor)
        self.up1 = Up(base_c * 16, base_c * 8 // factor, bilinear)
        self.up2 = Up(base_c * 8, base_c * 4 // factor, bilinear)
        self.up3 = Up(base_c * 4, base_c * 2 // factor, bilinear)
        self.up4 = Up(base_c * 2, base_c, bilinear)
        self.out_conv = OutConv(base_c, num_classes)

    def forward(self, x: torch.Tensor) -> Dict[str, torch.Tensor]:
        x1 = self.in_conv(x)
        x2 = self.down1(x1)
        x3 = self.down2(x2)
        x4 = self.down3(x3)
        x5 = self.down4(x4)
        x = self.up1(x5, x4)
        x = self.up2(x, x3)
        x = self.up3(x, x2)
        x = self.up4(x, x1) 
        logits = self.out_conv(x)
        return {"out":logits}

数据集

• test测试集

  • 1st_manual标签
  • 2nd_manual更加细致的标签
  • images原图
  • mask掩码，感兴趣的区域

• training训练集

  • 1st_manual
  • images
  • mask

• 数据处理部分

import os
from PIL import Image
import numpy as np 
from torch.utils.data import Dataset

class DriveDataset(Dataset):
    def __init__(self, root: str, train: bool, transforms=None):
        super(DriveDataset, self).__init__()
        self.flag = "training" if train else "test"
        data_root = os.path.join(root, "DRIVE", self.flag)
        assert os.path.exists(data_root), f"path '{data_root}' does not exist."
        self.transforms = transforms
        img_names = [i for i in os.listdir(os.path.join(data_root, "images")) if i.endswith(".tif")]
        self.img_list = [os.path.join(data_root, "images", i) for i in img_names]
        self.manual = [os.path.join(data_root, "1st_manual", i.split("_")[0] + "_manual1.gif")
                        for i in img_names]
        for i in self.manual:
            if not os.path.exists(i):
                raise FileNotFoundError(f"file {i} does not exist.")
        self.roi_mask = [os.path.join(data_root, "mask", i.split("_")[0] + f"_{self.flag}_mask.gif")
                        for i in img_names]
        for i in self.roi_mask:
            if not os.path.exists(i):
                raise FileNotFoundError(f"file {i} does not exist.")
        
    def __getitem__(self, idx):
        img = Image.open(self.img_list[idx]).convert("RGB")
        manual = Image.open(self.manual[idx]).convert("L")
        manual = np.array(manual) / 255
        roi_mask = Image.open(self.roi_mask[idx]).convert("L")  
        roi_mask = 255 - np.array(roi_mask)
        mask = np.clip(manual + roi_mask, a_min=0, a_max=255)
        mask = Image.fromarray(mask)
        if self.transforms is not None:
            img, mask = self.transforms(img, mask)
        return img, mask
    
    def __len__(self):
        return len(self.img_list)

    @staticmethod
    def collate_fn(batch):
        images, targets = list(zip(*batch))
        batched_imgs = DriveDataset.cat_list(images, fill_value=0)
        batched_targets = DriveDataset.cat_list(targets, fill_value=255)
        return batched_imgs, batched_targets

    @staticmethod
    def cat_list(images, fill_value=0):
        max_size = tuple(max(s) for s in zip(*[img.shape for img in images]))
        batch_shape = (len(images),) + max_size
        batched_imgs = images[0].new(*batch_shape).fill_(fill_value)
        for img, pad_img in zip(images, batched_imgs):
            pad_img[..., : img.shape[-2], : img.shape[-1]].copy_(img)
        return batched_imgs

• 代码文件不全，参考一下就行

transformer

整体网络架构

1. Embedding
   位置编码

• 它能为每个时间步输出一个独一无二的编码；
• 不同长度的句子之间，任何两个时间步之间的距离应该保持一致；
• 模型应该能毫不费力地泛化到更长的句子。
• 它的值应该是有界的；它必须是确定性的

注意力机制：从众多信息中选出对当前任务目标更加关键的信息

1. 基本的注意力机制
2. 在TRM中怎么操作

前馈神经网络

• 前馈网络（feed-forward network）是一种常见的神经网络结构，由一个或多个线性变换和非线性激活函数组成。它的输入是一个词向量，经过一系列线性变换和激活函数处理之后，输出另一个词向量。
• 前面都是线性变化（矩阵乘法），表达能力不够
• 更加深入的特征提取
  
  
  
  
  
  
  
  
  
  编码器：将输入文本序列编码为一系列隐藏表示，通常使用多层自注意力机制和前馈神经网络。
  解码器：接收编码器的输出和部分已生成的序列，使用自注意力机制和注意力机制来生成下一个词或字符。