Building a CNN model

梯度下降 https://towardsdatascience.com/gradient-descent-algorithm-a-deep-dive-cf04e8115f21
梯度下降仅适用于可微凹函数

CNN架构

构建CNN模型

1. 导入包
2. 收集数据
3. 准备数据
4. 构建模型
5. 训练模型
6. 评估模型的准确性

导入库

!pip install pytorch-lightning
!pip install opendatasets
import os
import shutil
import opendatasets as od
import pandas as pd
import numpy as np
from PIL import Image
from sklearn.metrics import confusion_matrix
from sklearn.model_selection import train_test_split
import matplotlib.pyplot as plt
import torch
from torch import nn, optim
from torch.utils.data import DataLoader, Dataset
from torch.utils.data.sampler import SubsetRandomSampler
from torchvision.datasets import ImageFolder
import torchvision.transforms as T
from torchvision.utils import make_grid
from torchmetrics.functional import accuracy
import pytorch_lightning as pl

收集数据

dataset_url = 'https://www.kaggle.com/c/histopathologic-cancer-detection'
od.download(dataset_url)
# 查看标签
cancer_labels = pd.read_csv('histopathologic-cancer-detection/train_labels.csv')
cancer_labels.head()
# 从train文件夹中随机选择了10000张图像作为数据集
np.random.seed(0)
train_imgs_orig = os.listdir("histopathologic-cancer-detection/train")
selected_image_list = []
for img in np.random.choice(train_imgs_orig, 10000):
  selected_image_list.append(img)
len(selected_image_list)
# 8000张训练集 2000张测试集
np.random.seed(0)
np.random.shuffle(selected_image_list)
cancer_train_idx = selected_image_list[:8000]
cancer_test_idx = selected_image_list[8000:]
print("Number of images in the downsampled training dataset: ", len(cancer_train_idx))
print("Number of images in the downsampled testing dataset: ", len(cancer_test_idx))

准备数据