原创申明： 未经许可，禁止以任何形式转载，若要引用，请标注链接地址
全文共计33000字，阅读大概需要5分钟
欢迎关注我的个人公众号：不懂开发的程序猿

一、配置环境

安装Anaconda

先检查电脑配置：NVIDIA RTX2060 + i7-10750H
我这里使用的是Windows操作系统 + 集成显卡：Inter® UHD Graphics + 独立显卡：NVIDIA RTX2060 ，以下实验基于此配置

安装的链接地址：Anacodan安装包官网地址https://repo.anaconda.com/archive/
这里我选择的是Anaconda3-5.2.0-Windows-x86_64，根据自己的操作系统选对应的版本，下载下来，右键管理员运行安装

Windows开始菜单管理员运行Anaconda Prompt控制台看到有base说明安装成功，输入conda --version

创建项目仓库，建议不要将项目的配置环境都装在base根目录仓库下，因为每个项目的配置环境不一样，如果项目运行是一样的环境就都用一样的解释器来运行，不一样的运行环境就建立不同的仓库，命令如下：
输入conda create -n pytorch python=3.6 创建pytorch仓库，选择y进行安装

输入：nvidia-smi检查显卡配置

安装pytorch

官网链接地址https://pytorch.org/
下拉网页，选择配置如下图，

运行命令：conda install pytorch torchvision torchaudio cudatoolkit=10.2 -c pytorch
选y确认安装，如下图。
安装过程大概10分钟

输入如下命令，没有报错，说明torch能正常调用显卡驱动

python
import torch
torch.cuda.is_available()

安装pyCharm

官网下载地址https://www.jetbrains.com/pycharm/download/#section=windows
傻瓜式安装即可，社区版和专业版都行，如果有edu邮箱的建议用一下专业版，白嫖它不香吗
新建项目如图：
Location：我选择的是E:\code\learn_pytorch
使用的虚拟环境是：E:\code\learn_pytorch\venv
Base 解释器：E:\Anaconda3-5.2.0\python.exe

测试pyCharm能否正确解析pytorch：在控制台输入

import torch
torch.cuda.is_available()

如图，没有报错，说明解析成功

安装jupyter

运行Anaconda Prompt控制台，输入

conda activate pytorch
conda install nb_conda

提示输入y，等待安装大概5分钟

输入：jupyter notebook

然后会跳转网页如图：

点击右边的new --> Python[conda env:pytorch]，新建一个终端

如果出现如下报错，删掉提示中的路径下的pythoncom36.dll文件即可

安装tensorboard

在Terminal中输入：

pip install tensorboard
tensorboard --logdir=logs --port=6007

安装opencv

pip install opencv-python

二、下载数据集

蚂蚁蜜蜂/练手数据集：链接:https://pan.baidu.com/s/1jZoTmoFzaTLWh4lKBHVbEA
密码: 5suq
解压下载好的数据集，粘贴在项目里
新建项目，编写测试代码

三、工具类的使用

tensorboard使用

# test_tensorboard.py
from torch.utils.tensorboard import SummaryWriter
import numpy as np
from PIL import Image

writer = SummaryWriter("logs")
image_path = "data/train/bees_image/16838648_415acd9e3f.jpg"
img_PIL = Image.open(image_path)
# 转换成numpy型数组
img_array = np.array(img_PIL)
print(type(img_array))
print(img_array.shape)

writer.add_image("test", img_array, 2, dataformats='HWC')
# y = x
for i in range(100):
    writer.add_scalar("y=2x", 3*i, i)

writer.close()

在Terminal中输入：
如果不–port=6007，就使用系统默认指定的6006端口

tensorboard --logdir=logs --port=6007

openCV读取图片

# read_data.py
from torch.utils.data import Dataset
#import cv2
from PIL import Image
import os


class MyData(Dataset):

    def __init__(self, root_dir, label_dir):
        # 申明全局变量
        self.root_dir = root_dir
        self.label_dir = label_dir
        # 获取路径
        self.path = os.path.join(self.root_dir,self.label_dir)
        #获取图片列表
        self.img_path = os.listdir(self.path)

    def __getitem__(self, idx):
        # 获取对应索引位置的图片的名字
        img_name = self.img_path[idx]
        img_item_path = os.path.join(self.root_dir, self.label_dir, img_name)
        # 读取图片
        img = Image.open(img_item_path)
        # 读取标签
        label = self.label_dir
        return img, label

    def __len__(self):
        return len(self.img_path)

# 读取数据集
root_dir = "dataset/train"
ants_label_dir = "ants"
bees_label_dir = "bees"
ants_dataset = MyData(root_dir, ants_label_dir)
bees_dataset = MyData(root_dir, bees_label_dir)

train_dataset = ants_dataset + bees_dataset

# rename_dataset.py
import os

root_dir = "dataset/train"
target_dir = "ants_image"
img_path = os.listdir(os.path.join(root_dir, target_dir))
label = target_dir.split('_')[0]
out_dir = "ants_label"
for i in img_path:
    file_name = i.split('_')[0]
    with open(os.path.join(root_dir, out_dir, "{}.txt".format(file_name)), 'w') as f:
        f.write(label)

Transforms使用

# useful_Transforms.py
from PIL import Image
from torch.utils.tensorboard import SummaryWriter
from torchvision import transforms

writer = SummaryWriter("logs")
img = Image.open("images/fb93a.jpeg")
print(img)

# ToTensor的使用
trans_totensor = transforms.ToTensor()
img_tensor = trans_totensor(img)
writer.add_image("ToTensor", img_tensor)

# Normalize() 归一化
# [channel] = (input[channel] - mean[channel]) / std[channel]
print(img_tensor[0][0][0])
trans_norma = transforms.Normalize([9, 6, 3], [1, 4, 7])
img_norma = trans_norma(img_tensor)
print(img_tensor[0][0][0])
writer.add_image("Normalize", img_norma, 2)

# Resize
print(img.size)
trans_resize = transforms.Resize((512, 512))
# img -> resize -> img_resize PIL
img_resize = trans_resize(img)
# img_resize PIL -> totensor -> img_resize tensor
img_resize = trans_totensor(img_resize)
writer.add_image("Resize", img_resize, 0)
print(img_resize)

# Compose - resize - 2
trans_resize_2 = transforms.Resize(512)
# PIL -> PIL -> tensor
trans_compose = transforms.Compose([trans_resize_2, trans_totensor])
img_resize_2 = trans_compose(img)
writer.add_image("Resize", img_resize_2, 1)

writer.close()

# Transforms.py
from PIL import Image
from torch.utils.tensorboard import SummaryWriter
from torchvision import transforms

# tensor的数据类型
# 通过transforms.ToTensor调用

# 2.为什么需要tensor数据类型

# 绝对路径 E:\code\learn_pytorch\data\train\ants_image\0013035.jpg Windows操作系统会把 \ 当做转义字符，不推荐使用绝对路径
# 相对路径 data/train/ants_image/0013035.jpg
img_path = "data/train/ants_image/0013035.jpg"
img = Image.open(img_path)

writer = SummaryWriter("logs")

# 1.transforms该如何使用
tensor_trans = transforms.ToTensor()
tensor_img = tensor_trans(img)

writer.add_image("Tensor_image", tensor_img)

writer.close()

CIFAR10公开数据集的使用

CIFAR-10是一个更接近普适物体的彩色图像数据集。CIFAR-10 是由Hinton 的学生Alex Krizhevsky 和Ilya Sutskever 整理的一个用于识别普适物体的小型数据集。一共包含10 个类别的RGB 彩色图片：飞机（ airplane ）、汽车（ automobile ）、鸟类（ bird ）、猫（ cat ）、鹿（ deer ）、狗（ dog ）、蛙类（ frog ）、马（ horse ）、船（ ship ）和卡车（ truck ）。
每个图片的尺寸为32 × 32 ，每个类别有6000个图像，数据集中一共有50000 张训练图片和10000 张测试图片。

# dataset_transforms.py
import torchvision
from torch.utils.tensorboard import SummaryWriter

dataset_transforms = torchvision.transforms.Compose([
    torchvision.transforms.ToTensor()
])

train_set = torchvision.datasets.CIFAR10(root="./dataset", train=True, transform=dataset_transforms, download=True)
test_set = torchvision.datasets.CIFAR10(root="./dataset", train=False, transform=dataset_transforms, download=True)

# print(test_set[0])
# print(test_set.classes)
#
# img, target = test_set[0]
# print(img)
# print(target)
# print(test_set.classes[3])
# img.show()

writer = SummaryWriter("logs")
for i in range(10):
    img, target = test_set[i]
    writer.add_image("test_set", img, i)

writer.close()

DataLoader的使用

# dataloader.py
import torchvision
from torch.utils.data import DataLoader

# 准备测试数据集
from torch.utils.tensorboard import SummaryWriter

test_data = torchvision.datasets.CIFAR10("./dataset", train=False, transform=torchvision.transforms.ToTensor())

test_loader = DataLoader(dataset=test_data, batch_size=64, shuffle=True, num_workers=0, drop_last=True)

#测试数据集中第一张图片及target
img, target = test_data[0]
print(img.shape)
print(target)

writer = SummaryWriter("logs")
for epoch in range(2):
    step =0
    for data in test_loader:
        imgs, targets = data
        writer.add_images("Epoch: {}".format(epoch), imgs, step)
        step += 1

writer.close()

shuffle=False	每次都是一样的取

shuffle=True	随机抽取

四、神经网络的搭建

# nn_module.py
# @Time: 2022/1/13 21:42
# @Author: 金阳
# @Parameter：
# @Version: 1.0.1

import torch
from torch import nn


class NeuralNetwork(nn.Module):

    def __init__(self) -> None:
        super().__init__()

    def forward(self, input):
        output = input + 1
        return output


neuralnetwork = NeuralNetwork()
x = torch.tensor(1.0)
output = neuralnetwork(x)
print(output)

二维卷积层Convolution Layers

# nn_conv.py
# @Time: 2022/1/13 22:08
# @Author: 金阳
# @Parameter：
# @Version: 1.0.1

import torch
import torch.nn.functional as F
input = torch.tensor([[1, 2, 0, 3, 1],
                      [0, 1, 2, 3, 1],
                      [1, 2, 1, 0, 0],
                      [5, 2, 3, 1, 1],
                      [2, 1, 0, 1, 1]])

kernel = torch.tensor([[1, 2, 1],
                       [0, 1, 0],
                       [2, 1, 0]])

input = torch.reshape(input, (1, 1, 5, 5))
kernel = torch.reshape(kernel, (1, 1, 3, 3))

print(input.shape)
print(kernel.shape)

output = F.conv2d(input, kernel, stride=1)
print(output)

output2 = F.conv2d(input, kernel, stride=2)
print(output2)

output3 = F.conv2d(input, kernel, stride=1, padding=1)
print(output3)

# src/nn_conv2d.py
# @Time: 2022/1/13 22:37
# @Author: 金阳
# @Parameter：
# @Version: 1.0.1

import torch
import torchvision
from torch import nn
from torch.nn import Conv2d
from torch.utils.data import DataLoader
from torch.utils.tensorboard import SummaryWriter

dataset = torchvision.datasets.CIFAR10("../dataset", train=False, transform=torchvision.transforms.ToTensor(),
                                       download=True)
dataloader = DataLoader(dataset, batch_size=64)

class NeuralNetwork(nn.Module):
    def __init__(self):
        super(NeuralNetwork, self).__init__()
        self.conv1 = Conv2d(in_channels=3, out_channels=6, kernel_size=3, stride=1, padding=0)

    def forward(self,x):
        x = self.conv1(x)
        return x

neuralnetwork = NeuralNetwork()
# print(neuralnetwork)

writer = SummaryWriter("../logs")
step = 0
for data in dataloader:
    imgs, targets = data
    output = neuralnetwork(imgs)
    print(imgs.shape)
    print(output.shape)

    # torch.Size([64, 3, 32, 32])
    writer.add_images("input", imgs, step)

    # torch.Size([64, 6, 30, 30])
    output = torch.reshape(output, (-1, 3, 30, 30))
    writer.add_images("output", output, step)
    step += 1

writer.close()

池化层 Pooling layers

# nn_maxpool.py
# @Time: 2022/1/14 9:53
# @Author: 金阳
# @Parameter：
# @Version: 1.0.1

import torch
from torch import nn
from torch.nn import MaxPool2d

input = torch.tensor([[1, 2, 0, 3, 1],
                      [0, 1, 2, 3, 1],
                      [1, 2, 1, 0, 0],
                      [5, 2, 3, 1, 1],
                      [2, 1, 0, 1, 1]], dtype=torch.float32)

input = torch.reshape(input, (-1, 1, 5, 5))
print(input.shape)

class NueralNetwork(nn.Module):
    def __init__(self):
        super(NueralNetwork, self).__init__()
        self.maxpool1 = MaxPool2d(kernel_size=3, ceil_mode=True)

    def forward(self, input):
        output = self.maxpool1(input)
        return output

nueralnetwork = NueralNetwork()
output = nueralnetwork(input)

print(output)
tensor([[[[2., 3.],
          [5., 1.]]]])

ceil_mode=False
tensor([[[[2.]]]])

# @Time: 2022/1/14 9:53
# @Author: 金阳
# @Parameter：
# @Version: 1.0.1

import torch
import torchvision
from torch import nn
from torch.nn import MaxPool2d
from torch.utils.data import DataLoader
from torch.utils.tensorboard import SummaryWriter

dataset = torchvision.datasets.CIFAR10("../dataset", train=False, transform=torchvision.transforms.ToTensor(),
                                       download=True)
dataloader = DataLoader(dataset, batch_size=64)


# input = torch.tensor([[1, 2, 0, 3, 1],
#                       [0, 1, 2, 3, 1],
#                       [1, 2, 1, 0, 0],
#                       [5, 2, 3, 1, 1],
#                       [2, 1, 0, 1, 1]], dtype=torch.float32)
#
# input = torch.reshape(input, (-1, 1, 5, 5))
# print(input.shape)

class NueralNetwork(nn.Module):
    def __init__(self):
        super(NueralNetwork, self).__init__()
        self.maxpool1 = MaxPool2d(kernel_size=3, ceil_mode=True)

    def forward(self, input):
        output = self.maxpool1(input)
        return output

nueralnetwork = NueralNetwork()

writer = SummaryWriter("../logs")
step = 0
for data in dataloader:
    imgs, targets = data
    writer.add_images("input", imgs, step)
    output = nueralnetwork(imgs)
    writer.add_images("output", output, step)
    step += 1

writer.close()

# output = nueralnetwork(input)
#
# print(output)

非线性激活Non-linear Activations (weighted sum, nonlinearity)

nn. ReLU

import torch
import torchvision
from torch import nn
from torch.nn import ReLU
from torch.utils.data import DataLoader

input = torch.tensor([[1, -0.5],
                      [-1, 3]])
input = torch.reshape(input, (-1, 1, 2, 2))
print(input.shape)
print(input)
# tensor([[[[ 1.0000, -0.5000],
          [-1.0000,  3.0000]]]])

class NerualNetwork(nn.Module):
    def __init__(self):
        super(NerualNetwork, self).__init__()
        self.relu1 = ReLU()

    def forward(self, input):
        output = self.relu1(input)
        return output

nerualnetwork = NerualNetwork()
output = nerualnetwork(input)
print(output)
tensor([[[[1., 0.],
          [0., 3.]]]])

nn.Sigmoid

tensorboard里面 step 不从0 开始，是图片显示的问题， 用命令：

tensorboard --logdir=logs --samples_per_plugin images=1000

# @Time: 2022/1/14 10:34
# @Author: 金阳
# @Parameter：
# @Version: 1.0.1

import torch
import torchvision
from torch import nn
from torch.nn import ReLU, Sigmoid
from torch.utils.data import DataLoader
from torch.utils.tensorboard import SummaryWriter

input = torch.tensor([[1, -0.5],
                      [-1, 3]])
input = torch.reshape(input, (-1, 1, 2, 2))
print(input.shape)
print(input)

dataset = torchvision.datasets.CIFAR10("../dataset", train=False, transform=torchvision.transforms.ToTensor(),
                                       download=True)
dataloader = DataLoader(dataset, batch_size=64, )

class NerualNetwork(nn.Module):
    def __init__(self):
        super(NerualNetwork, self).__init__()
        self.relu1 = ReLU()
        self.sigmod1 = Sigmoid()

    def forward(self, input):
        output = self.sigmod1(input)
        return output

nerualnetwork = NerualNetwork()

writer = SummaryWriter("../logs")
step = 0
for data in dataloader:
    imgs, target = data
    writer.add_images("input", imgs, global_step=step)
    output = nerualnetwork(imgs)
    writer.add_images("output", output, global_step=step)
    step += 1

writer.close()

正则化Normalization Layers

nn.BatchNorm2d

Batch Normalization: Accelerating Deep Network Training by Reducing Internal Covariate Shift .

Recurrent Layers

nn.RNN

nn.LSTM

Sparse Layers

nn.Embedding

A simple lookup table that stores embeddings of a fixed dictionary and size.

Linear Layers

# nn_linear.py
# @Time: 2022/1/14 11:48
# @Author: 金阳
# @Parameter：
# @Version: 1.0.1
import torch
import torchvision
from torch import nn
from torch.nn import Linear
from torch.utils.data import DataLoader

dataset = torchvision.datasets.CIFAR10("../dataset", train=False, transform=torchvision.transforms.ToTensor(),
                                       download=True)
dataloader = DataLoader(dataset, batch_size=64)

class NerualNetwork(nn.Module):
    def __init__(self):
        super(NerualNetwork, self).__init__()
        self.linear1 = Linear(196608, 10)

    def forward(self, input):
        output = self.linear1(input)
        return output

nerualnetwork = NerualNetwork()


for data in dataloader:
    imgs, targets =data
    print(imgs.shape)
    # torch.Size([64, 3, 32, 32])
    # output = torch.reshape(imgs, (1, 1, 1, -1))
    output = torch.flatten(imgs)
    print(output.shape)
    # torch.Size([196608])
    output = nerualnetwork(output)
    print(output.shape)
    # torch.Size([10])

SEQUENTIAL

# nn_seq.py
# @Time: 2022/1/14 12:57
# @Author: 金阳
# @Parameter：
# @Version: 1.0.1
import torch
from torch import nn
from torch.nn import Conv2d, MaxPool2d, Flatten, Linear, Sequential
from torch.utils.tensorboard import SummaryWriter


class NueralNetwork(nn.Module):
    def __init__(self):
        super(NueralNetwork, self).__init__()
        # self.conv1 = Conv2d(in_channels=3, out_channels=32, kernel_size=5, padding=2)
        # self.maxpool1 = MaxPool2d(kernel_size=2)
        # self.conv2 = Conv2d(in_channels=32, out_channels=32, kernel_size=5, padding=2)
        # self.maxpool2 = MaxPool2d(kernel_size=2)
        # self.conv3 = Conv2d(in_channels=32, out_channels=64, kernel_size=5, padding=2)
        # self.maxpool3 = MaxPool2d(kernel_size=2)
        # self.flatten = Flatten()
        # self.linear1 = Linear(in_features=1024, out_features=64)
        # self.linear2 = Linear(in_features=64, out_features=10)

        # 简单的写法
        self.model1 = Sequential(
            Conv2d(in_channels=3, out_channels=32, kernel_size=5, padding=2),
            MaxPool2d(kernel_size=2),
            Conv2d(in_channels=32, out_channels=32, kernel_size=5, padding=2),
            MaxPool2d(kernel_size=2),
            Conv2d(in_channels=32, out_channels=64, kernel_size=5, padding=2),
            MaxPool2d(kernel_size=2),
            Flatten(),
            Linear(in_features=1024, out_features=64),
            Linear(in_features=64, out_features=10)
        )

    def forward(self, x):
        # x = self.conv1(x)
        # x = self.maxpool1(x)
        # x = self.conv2(x)
        # x = self.maxpool2(x)
        # x = self.conv3(x)
        # x = self.maxpool3(x)
        # x = self.flatten(x)
        # x = self.linear1(x)
        # x = self.linear2(x)

        x = self.model1(x)

        return x

nueralnetwork = NueralNetwork()
print(nueralnetwork)

input = torch.ones((64, 3, 32, 32))
output = nueralnetwork(input)
print(output.shape)


writer = SummaryWriter("../logs")
writer.add_graph(nueralnetwork, input)
writer.close()

输出结果是：

NueralNetwork(
  (model1): Sequential(
    (0): Conv2d(3, 32, kernel_size=(5, 5), stride=(1, 1), padding=(2, 2))
    (1): MaxPool2d(kernel_size=2, stride=2, padding=0, dilation=1, ceil_mode=False)
    (2): Conv2d(32, 32, kernel_size=(5, 5), stride=(1, 1), padding=(2, 2))
    (3): MaxPool2d(kernel_size=2, stride=2, padding=0, dilation=1, ceil_mode=False)
    (4): Conv2d(32, 64, kernel_size=(5, 5), stride=(1, 1), padding=(2, 2))
    (5): MaxPool2d(kernel_size=2, stride=2, padding=0, dilation=1, ceil_mode=False)
    (6): Flatten(start_dim=1, end_dim=-1)
    (7): Linear(in_features=1024, out_features=64, bias=True)
    (8): Linear(in_features=64, out_features=10, bias=True)
  )
)
torch.Size([64, 10])

损失函数Loss Functions

nn.L1Loss

nn.MSELoss

# nn_loss_network.py
import torchvision
from torch import nn
from torch.nn import Conv2d, MaxPool2d, Flatten, Linear, Sequential
from torch.utils.data import DataLoader

dataset = torchvision.datasets.CIFAR10("../dataset", train=False, download=True,
                                       transform=torchvision.transforms.ToTensor())
dataloader = DataLoader(dataset, batch_size=1)

class NueralNetwork(nn.Module):
    def __init__(self):
        super(NueralNetwork, self).__init__()
        self.model1 = Sequential(
            Conv2d(in_channels=3, out_channels=32, kernel_size=5, padding=2),
            MaxPool2d(kernel_size=2),
            Conv2d(in_channels=32, out_channels=32, kernel_size=5, padding=2),
            MaxPool2d(kernel_size=2),
            Conv2d(in_channels=32, out_channels=64, kernel_size=5, padding=2),
            MaxPool2d(kernel_size=2),
            Flatten(),
            Linear(in_features=1024, out_features=64),
            Linear(in_features=64, out_features=10)
        )

    def forward(self, x):
        x = self.model1(x)
        return x

loss = nn.CrossEntropyLoss()
nueralnetwork = NueralNetwork()
for data in dataloader:
    imgs, targets = data
    outputs = nueralnetwork(imgs)
    result_loss = loss(outputs, targets)
    print(result_loss)
    print(outputs)
    print(targets)

nn.CrossEntropyLoss

优化器TORCH.OPTIM

# nn_optim.py
# @Time: 2022/1/14 20:21
# @Author: 金阳
# @Parameter：
# @Version: 1.0.1
import torch
import torchvision
from torch import nn
from torch.nn import Conv2d, MaxPool2d, Flatten, Linear, Sequential
from torch.utils.data import DataLoader

dataset = torchvision.datasets.CIFAR10("../dataset", train=False, download=True,
                                       transform=torchvision.transforms.ToTensor())
dataloader = DataLoader(dataset, batch_size=1)

class NueralNetwork(nn.Module):
    def __init__(self):
        super(NueralNetwork, self).__init__()
        self.model1 = Sequential(
            Conv2d(in_channels=3, out_channels=32, kernel_size=5, padding=2),
            MaxPool2d(kernel_size=2),
            Conv2d(in_channels=32, out_channels=32, kernel_size=5, padding=2),
            MaxPool2d(kernel_size=2),
            Conv2d(in_channels=32, out_channels=64, kernel_size=5, padding=2),
            MaxPool2d(kernel_size=2),
            Flatten(),
            Linear(in_features=1024, out_features=64),
            Linear(in_features=64, out_features=10)
        )

    def forward(self, x):
        x = self.model1(x)
        return x

loss = nn.CrossEntropyLoss()
nueralnetwork = NueralNetwork()
# 定义优化器
optim = torch.optim.SGD(nueralnetwork.parameters(), lr=0.01)
for epoch in range(20):
    running_loss = 0.0
    for data in dataloader:
        imgs, targets = data
        outputs = nueralnetwork(imgs)
        result_loss = loss(outputs, targets)
        # 梯度清零
        optim.zero_grad()
        result_loss.backward()
        optim.step()
        running_loss += result_loss
    print(running_loss)

断点调试

网络模型的增删改

# model_pretrained.py
import torchvision

# train_data = torchvision.datasets.ImageNet("../data_image_net", split='train', download=True,
#                                            transform=torchvision.transforms.ToTensor())
RuntimeError: The dataset is no longer publicly accessible. You need to download the archives externally and place them in the root directory.
from torch import nn

vgg16_false = torchvision.models.vgg16(pretrained=False)
vgg16_true = torchvision.models.vgg16(pretrained=True)

print(vgg16_true)

train_data = torchvision.datasets.CIFAR10('../data', train=True, transform=torchvision.transforms.ToTensor(),
                                          download=True)

vgg16_true.classifier.add_module('add_linear', nn.Linear(1000, 10))
print(vgg16_true)

print(vgg16_false) 

vgg16_false.classifier[6] = nn.Linear(4096, 10)
print(vgg16_false)

数据集下载的位置：Downloading:
https://download.pytorch.org/models/vgg16-397923af.pth
to C:\Users\15718/.cache\torch\hub\checkpoints\vgg16-397923af.pth

***如果遇到数据集较大，网络不通畅，建议：复制下载链接，用迅雷下 ***

模型的保存与加载

# model_save.py
import torch
import torchvision
from torch import nn

vgg16 = torchvision.models.vgg16(pretrained=False)
# 保存方式一     模型结构 + 模型参数
torch.save(vgg16, "vgg16_method1.pth")

# 保存方式二     模型参数（官方推荐）
torch.save(vgg16.state_dict(), "vgg16_method2.pth")

# 陷阱
class NeuralNetwork(nn.Module):
    def __init__(self):
        super(NeuralNetwork, self).__init__()
        self.conv1 = nn.Conv2d(in_channels=3, out_channels=6, kernel_size=3, stride=1, padding=0)

    def forward(self,x):
        x = self.conv1(x)
        return x

neuralnetwork = NeuralNetwork()
torch.save(neuralnetwork, "neuralnetwork_method1.pth")

# model_load.py
import torch
from model_save import *

# 方式一  --》 保存方式一，加载模型
model = torch.load("vgg16_method1.pth")
# print(model)

# 方式二   加载模型，字典格式的
model = torch.load("vgg16_method2.pth")
print(model)

# 方式二   加载模型，VGG格式的
vgg16 = torchvision.models.vgg16(pretrained=False)
vgg16.load_state_dict(torch.load("vgg16_method2.pth"))
print(vgg16)

# 陷阱
model = torch.load("neuralnetwork_method1.pth")
print(model)

模型训练套路

# model.py
# @Time: 2022/1/15 9:03
# @Author: 金阳
# @Parameter：
# @Version: 1.0.1
import torch
from torch import nn

#搭建神经网络
class NurealNetwork(nn.Module):
    def __init__(self):
        super(NurealNetwork, self).__init__()
        self.model = nn.Sequential(
            nn.Conv2d(in_channels=3, out_channels=32, kernel_size=5, stride=1, padding=2),
            nn.MaxPool2d(kernel_size=2),
            nn.Conv2d(in_channels=32, out_channels=32, kernel_size=5, stride=1, padding=2),
            nn.MaxPool2d(kernel_size=2),
            nn.Conv2d(in_channels=32, out_channels=64, kernel_size=5, stride=1, padding=2),
            nn.MaxPool2d(kernel_size=2),
            nn.Flatten(),
            nn.Linear(64*4*4, 64),
            nn.Linear(64, 10)
        )

    def forward(self, x):
        x = self.model(x)
        return x

if __name__ == '__main__':
    nurealnetwork = NurealNetwork()
    input = torch.ones((64, 3, 32, 32))
    output = nurealnetwork(input)
    print(output.shape)

# train.py
# @Time: 2022/1/15 8:43
# @Author: 金阳
# @Parameter：
# @Version: 1.0.1

# 准备数据集
import torch.optim
import torchvision
from torch.utils.data import DataLoader
from torch.utils.tensorboard import SummaryWriter

from model import *

train_data = torchvision.datasets.CIFAR10(root="../dataset", train=True,transform=torchvision.transforms.ToTensor(),
                                          download=True)
test_data = torchvision.datasets.CIFAR10(root="../dataset", train=False, transform=torchvision.transforms.ToTensor(),
                                         download=True)

train_data_size = len(train_data)
test_data_size = len(test_data)
# python中格式化字符串的写法，如果train_data_size = 10, 输出-->训练数据集的长度为：10
print("训练数据集的长度为：{}".format(train_data_size))
print("测试数据集的长度为：{}".format(test_data_size))

# 利用DataLoader加载数据集
train_dataloader = DataLoader(train_data,batch_size=64)
test_dataloader = DataLoader(test_data, batch_size=64)

# 创建神经网络
nurealnetwork = NurealNetwork()

# 损失函数
loss_fn = nn.CrossEntropyLoss()

# 优化器
# learning_rate = 0.01
learning_rate = 1e-2
optimizer = torch.optim.SGD(nurealnetwork.parameters(), lr=learning_rate)

# 设置训练网络的一些参数
# 记录训练次数
total_train_step = 0
# 记录测试次数
total_test_step = 0
# 训练的轮数
epoch = 10

# 添加Tensorboard
writer = SummaryWriter("../logs_train")

for i in range(epoch):
    print("------第 {} 轮训练开始------".format(i + 1))


    # 训练开始
    nurealnetwork.train()
    for data in train_dataloader:
        imgs, targets = data
        outputs = nurealnetwork(imgs)
        loss = loss_fn(outputs, targets)

        # 优化器优化模型
        # 梯度清零
        optimizer.zero_grad()
        # 反向传播得出每一个梯度
        loss.backward()
        # 对其中的参数进行优化
        optimizer.step()
        # 训练结束，total_train_step加一
        total_train_step += 1
        # 减少打印的量
        if total_train_step % 100 == 0:
            print("训练次数：{} , Loss: {}".format(total_train_step, loss.item()))
            writer.add_scalar("train_loss", loss.item(), total_train_step)

    #验证集
    nurealnetwork.eval()
    total_test_loss = 0
    total_accuracy = 0
    with torch.no_grad():
        for data in test_dataloader:
            imgs, targets = data
            outputs = nurealnetwork(imgs)
            # 比较输出与目标之间的差距
            loss = loss_fn(outputs, targets)
            total_test_loss += loss.item()
            accuracy = (outputs.argmax(1) == targets).sum()
            total_accuracy += accuracy
    print("整体测试集上的Loss：{}".format(total_test_loss))
    print("整体测试集上的正确率：{}".format(total_accuracy / test_data_size))
    writer.add_scalar("test_loss", total_test_loss, total_test_step)
    writer.add_scalar("test_accracy", total_accuracy / test_data_size, total_test_step)
    total_test_step += 1

    # 保存模型
    torch.save(nurealnetwork, "nurealnetwork_{}.pth".format(i))
    #官方推荐保存方式
    # torch.save(nurealnetwork.state_dict(), "nurealnetwork_{}.pth".format(i))
    print("模型已保存")

writer.close()

GPU训练模型

CPU训练 vs GPU训练

CPU 用时

GPU 用时

用任意的模型测试神经网络能否预测成功

# test.py

# @Time: 2022/1/15 15:40
# @Author: 金阳
# @Parameter：
# @Version: 1.0.1
import torch
import torchvision
from PIL import Image
from torch import nn

image_path = "../images/dog.png"
image = Image.open(image_path)
print(image)
image = image.convert("RGB")

transform =torchvision.transforms.Compose([torchvision.transforms.Resize((32, 32)),
                                           torchvision.transforms.ToTensor()])
image =transform(image)
print(image.shape)

class Tudui(nn.Module):
    def __init__(self):
        super(Tudui, self).__init__()
        self.model = nn.Sequential(
            nn.Conv2d(in_channels=3, out_channels=32, kernel_size=5, stride=1, padding=2),
            nn.MaxPool2d(kernel_size=2),
            nn.Conv2d(in_channels=32, out_channels=32, kernel_size=5, stride=1, padding=2),
            nn.MaxPool2d(kernel_size=2),
            nn.Conv2d(in_channels=32, out_channels=64, kernel_size=5, stride=1, padding=2),
            nn.MaxPool2d(kernel_size=2),
            nn.Flatten(),
            nn.Linear(64*4*4, 64),
            nn.Linear(64, 10)
        )

    def forward(self, x):
        x = self.model(x)
        return x
model = torch.load("tudui_29_gpu.pth")
print(model)
image = torch.reshape(image, (1, 3, 32, 32))
model.eval()
with torch.no_grad():
    image = image.cuda()
    output = model(image)
print(output)

print(output.argmax(1))

五、项目实战

YOLO v5训练

下载代码：https://github.com/ultralytics/yolov5/tree/v6.0

安装，配置环境

pip install -r requirements.txt

如果安装依赖库太慢了或者失败了，建议在Anaconda命令行里，进入项目的对应仓库里，安装依赖库，命令如下

pip install 依赖库 -i https://pypi.tuna.tsinghua.edu.cn/simple

def parse_opt():
    parser = argparse.ArgumentParser()
    # 设置权重，可选预训练数据集模型有yolov5s.pt yolov5m.pt yolov5l.pt yolov5x.pt 依次增大，训练所需时间也相应更久，效果更好
    parser.add_argument('--weights', nargs='+', type=str, default=ROOT / 'yolov5s.pt', help='model path(s)')
    # 设置训练的径在哪里
    parser.add_argument('--source', type=str, default=ROOT / 'data/images', help='file/dir/URL/glob, 0 for webcam')
    # 默认指定了图片尺寸640px，
    parser.add_argument('--imgsz', '--img', '--img-size', nargs='+', type=int, default=[640], help='inference size h,w')
    parser.add_argument('--conf-thres', type=float, default=0.25, help='confidence threshold')
    parser.add_argument('--iou-thres', type=float, default=0.45, help='NMS IoU threshold')
    parser.add_argument('--max-det', type=int, default=1000, help='maximum detections per image')
    #指定是用CPU训练还是GPU训练，如果有多个GPU，可以同时开启
    parser.add_argument('--device', default='', help='cuda device, i.e. 0 or 0,1,2,3 or cpu')
    # 训练过程中可以预览图片，如果是训练视频则可以预览视频
    parser.add_argument('--view-img', action='store_true', help='show results')
    parser.add_argument('--save-txt', action='store_true', help='save results to *.txt')
    parser.add_argument('--save-conf', action='store_true', help='save confidences in --save-txt labels')
    parser.add_argument('--save-crop', action='store_true', help='save cropped prediction boxes')
    parser.add_argument('--nosave', action='store_true', help='do not save images/videos')
    parser.add_argument('--classes', nargs='+', type=int, help='filter by class: --classes 0, or --classes 0 2 3')
    parser.add_argument('--agnostic-nms', action='store_true', help='class-agnostic NMS')
    parser.add_argument('--augment', action='store_true', help='augmented inference')
    parser.add_argument('--visualize', action='store_true', help='visualize features')
    parser.add_argument('--update', action='store_true', help='update all models')
    parser.add_argument('--project', default=ROOT / 'runs/detect', help='save results to project/name')
    parser.add_argument('--name', default='exp', help='save results to project/name')
    parser.add_argument('--exist-ok', action='store_true', help='existing project/name ok, do not increment')
    parser.add_argument('--line-thickness', default=3, type=int, help='bounding box thickness (pixels)')
    parser.add_argument('--hide-labels', default=False, action='store_true', help='hide labels')
    parser.add_argument('--hide-conf', default=False, action='store_true', help='hide confidences')
    parser.add_argument('--half', action='store_true', help='use FP16 half-precision inference')
    parser.add_argument('--dnn', action='store_true', help='use OpenCV DNN for ONNX inference')
    opt = parser.parse_args()
    opt.imgsz *= 2 if len(opt.imgsz) == 1 else 1  # expand
    print_args(FILE.stem, opt)
    return opt

效果如图：

视频检测

将batch_size = 32 调为4，不然显卡会带不动
测试集跑完整个视频，大概花了半个小时
本地跑不动的可以用Google colab 云端服务器跑

if __name__ == '__main__':
    parser = argparse.ArgumentParser()
    parser.add_argument('--weights', nargs='+', type=str, default='yolov5s.pt', help='model.pt path(s)')
    # 指定视频的路径
    parser.add_argument('--source', type=str, default='data/video/movie.mp4', help='source')  # file/folder, 0 for webcam
    parser.add_argument('--img-size', type=int, default=640, help='inference size (pixels)')
    parser.add_argument('--conf-thres', type=float, default=0.25, help='object confidence threshold')
    parser.add_argument('--iou-thres', type=float, default=0.45, help='IOU threshold for NMS')
    parser.add_argument('--device', default='', help='cuda device, i.e. 0 or 0,1,2,3 or cpu')
    parser.add_argument('--view-img', action='store_true', help='display results')
    parser.add_argument('--save-txt', action='store_true', help='save results to *.txt')
    parser.add_argument('--save-conf', action='store_true', help='save confidences in --save-txt labels')
    parser.add_argument('--nosave', action='store_true', help='do not save images/videos')
    parser.add_argument('--classes', nargs='+', type=int, help='filter by class: --class 0, or --class 0 2 3')
    parser.add_argument('--agnostic-nms', action='store_true', help='class-agnostic NMS')
    parser.add_argument('--augment', action='store_true', help='augmented inference')
    parser.add_argument('--update', action='store_true', help='update all models')
    parser.add_argument('--project', default='runs/detect', help='save results to project/name')
    parser.add_argument('--name', default='exp', help='save results to project/name')
    parser.add_argument('--exist-ok', action='store_true', help='existing project/name ok, do not increment')
    opt = parser.parse_args()
    print(opt)
    check_requirements(exclude=('pycocotools', 'thop'))

    with torch.no_grad():
        if opt.update:  # update all models (to fix SourceChangeWarning)
            for opt.weights in ['yolov5s.pt', 'yolov5m.pt', 'yolov5l.pt', 'yolov5x.pt']:
                detect()
                strip_optimizer(opt.weights)
        else:
            detect()

视频截图：

训练coco128数据集

设置参数

训练结果：

VOC2007(Visual Object Classes)数据集

Annotations：包含了xml文件，描述了图片的各种信息，特别是标注出了目标的位置坐标
ImageSets：主要关注下main文件夹的内容，里面的文件包含了不同类别的训练/验证数据集图片名称
JPEGImages：原图片
SegmentatioClass、SegmenObject：语义分割

coco数据集(Common Objects in Context)

待更新

自制数据集，训练

待更新

FaceMaskDetection 人脸面罩检测

源码https://github.com/AIZOOTech/FaceMaskDetection

# -*- coding:utf-8 -*-
import cv2
import time

import argparse
import numpy as np
from PIL import Image
from utils.anchor_generator import generate_anchors
from utils.anchor_decode import decode_bbox
from utils.nms import single_class_non_max_suppression
from load_model.pytorch_loader import load_pytorch_model, pytorch_inference

# model = load_pytorch_model('models/face_mask_detection.pth');
model = load_pytorch_model('models/model360.pth')
# anchor configuration
#feature_map_sizes = [[33, 33], [17, 17], [9, 9], [5, 5], [3, 3]]
feature_map_sizes = [[45, 45], [23, 23], [12, 12], [6, 6], [4, 4]]
anchor_sizes = [[0.04, 0.056], [0.08, 0.11], [0.16, 0.22], [0.32, 0.45], [0.64, 0.72]]
anchor_ratios = [[1, 0.62, 0.42]] * 5

# generate anchors
anchors = generate_anchors(feature_map_sizes, anchor_sizes, anchor_ratios)

# for inference , the batch size is 1, the model output shape is [1, N, 4],
# so we expand dim for anchors to [1, anchor_num, 4]
anchors_exp = np.expand_dims(anchors, axis=0)

id2class = {0: 'Mask', 1: 'NoMask'}


def inference(image,
              conf_thresh=0.5,
              iou_thresh=0.4,
              target_shape=(160, 160),
              draw_result=True,
              show_result=True
              ):
    '''
    Main function of detection inference
    :param image: 3D numpy array of image
    :param conf_thresh: the min threshold of classification probabity.
    :param iou_thresh: the IOU threshold of NMS
    :param target_shape: the model input size.
    :param draw_result: whether to daw bounding box to the image.
    :param show_result: whether to display the image.
    :return:
    '''
    # image = np.copy(image)
    output_info = []
    height, width, _ = image.shape
    image_resized = cv2.resize(image, target_shape)
    image_np = image_resized / 255.0  # 归一化到0~1
    image_exp = np.expand_dims(image_np, axis=0)

    image_transposed = image_exp.transpose((0, 3, 1, 2))

    y_bboxes_output, y_cls_output = pytorch_inference(model, image_transposed)
    # remove the batch dimension, for batch is always 1 for inference.
    y_bboxes = decode_bbox(anchors_exp, y_bboxes_output)[0]
    y_cls = y_cls_output[0]
    # To speed up, do single class NMS, not multiple classes NMS.
    bbox_max_scores = np.max(y_cls, axis=1)
    bbox_max_score_classes = np.argmax(y_cls, axis=1)

    # keep_idx is the alive bounding box after nms.
    keep_idxs = single_class_non_max_suppression(y_bboxes,
                                                 bbox_max_scores,
                                                 conf_thresh=conf_thresh,
                                                 iou_thresh=iou_thresh,
                                                 )

    for idx in keep_idxs:
        conf = float(bbox_max_scores[idx])
        class_id = bbox_max_score_classes[idx]
        bbox = y_bboxes[idx]
        # clip the coordinate, avoid the value exceed the image boundary.
        xmin = max(0, int(bbox[0] * width))
        ymin = max(0, int(bbox[1] * height))
        xmax = min(int(bbox[2] * width), width)
        ymax = min(int(bbox[3] * height), height)

        if draw_result:
            if class_id == 0:
                color = (0, 255, 0)
            else:
                color = (255, 0, 0)
            cv2.rectangle(image, (xmin, ymin), (xmax, ymax), color, 2)
            cv2.putText(image, "%s: %.2f" % (id2class[class_id], conf), (xmin + 2, ymin - 2),
                        cv2.FONT_HERSHEY_SIMPLEX, 0.8, color)
        output_info.append([class_id, conf, xmin, ymin, xmax, ymax])

    if show_result:
        Image.fromarray(image).show()
    return output_info


def run_on_video(video_path, output_video_name, conf_thresh):
    cap = cv2.VideoCapture(video_path)
    height = cap.get(cv2.CAP_PROP_FRAME_HEIGHT)
    width = cap.get(cv2.CAP_PROP_FRAME_WIDTH)
    fps = cap.get(cv2.CAP_PROP_FPS)
    fourcc = cv2.VideoWriter_fourcc(*'XVID')
    # writer = cv2.VideoWriter(output_video_name, fourcc, int(fps), (int(width), int(height)))
    total_frames = cap.get(cv2.CAP_PROP_FRAME_COUNT)
    if not cap.isOpened():
        raise ValueError("Video open failed.")
        return
    status = True
    idx = 0
    while status:
        start_stamp = time.time()
        status, img_raw = cap.read()
        img_raw = cv2.cvtColor(img_raw, cv2.COLOR_BGR2RGB)
        read_frame_stamp = time.time()
        if (status):
            inference(img_raw,
                      conf_thresh,
                      iou_thresh=0.5,
                      target_shape=(360, 360),
                      draw_result=True,
                      show_result=False)
            cv2.imshow('image', img_raw[:, :, ::-1])
            cv2.waitKey(1)
            inference_stamp = time.time()
            # writer.write(img_raw)
            write_frame_stamp = time.time()
            idx += 1
            print("%d of %d" % (idx, total_frames))
            print("read_frame:%f, infer time:%f, write time:%f" % (read_frame_stamp - start_stamp,
                                                                   inference_stamp - read_frame_stamp,
                                                                   write_frame_stamp - inference_stamp))
    # writer.release()


if __name__ == "__main__":
    parser = argparse.ArgumentParser(description="Face Mask Detection")
    parser.add_argument('--img-mode', type=int, default=1, help='set 1 to run on image, 0 to run on video.')
    parser.add_argument('--img-path', type=str, default='img_test/test5.jpg', help='path to your image.')
    parser.add_argument('--video-path', type=str, default='0', help='path to your video, `0` means to use camera.')
    # parser.add_argument('--hdf5', type=str, help='keras hdf5 file')
    args = parser.parse_args()
    if args.img_mode:
        imgPath = args.img_path
        img = cv2.imread(imgPath)
        img = cv2.cvtColor(img, cv2.COLOR_BGR2RGB)
        inference(img, show_result=True, target_shape=(360, 360))
    else:
        video_path = args.video_path
        if args.video_path == '0':
            video_path = 0
        run_on_video(video_path, '', conf_thresh=0.5)

图片检测结果：

python pytorch_infer.py  --img-path /path/to/your/img

前：

后：

视频流检测

python pytorch_infer.py --img-mode 0 --video-path /path/to/video  
# 如果要打开本地摄像头, video_path填写0就可以了，如下
python pytorch_infer.py --img-mode 0 --video-path 0