3.3. 采用Resnet18实现裂缝分类#
3.2节我们学习了ResNet18的网络架构,同时我们在第二章也对卷积、池化和激活函数的基础知识进行了详细描述。至此,我们已经具备了将ResNet18带入项目实例,实现对裂缝图像分类学习的全部基础知识。下面我们将依照深度学习的实现流程,实现三个场景下有无裂缝的项目进行分类(共计6个种类),具体流程如图3-7所示。
图3-7使用PyTorch进行深度学习的实现流程#
chapter_3_3_01.py#
1import os
2import random
3import datetime
4
5import numpy as np
6from PIL import Image
7from sklearn import model_selection
8import torch
9import torch.optim as optim
10from torch.optim import lr_scheduler
11from torch.utils import tensorboard
12from torch import nn, functional as F
13from torch.utils import data as torch_data
14import torchvision.models as models
15import torchvision.transforms as transforms