GoogLeNet是在2014年由Google团队提出的，获得了当年ImageNet比赛中分类任务的第一名，也就是和VGG是同一年提出的，在ImageNet比赛中都获得了很好的成绩。GoogLeNet的网络结构比较复杂，具体的结构可以参考原论文，论文名字是：Going Deeper with Convolutions。
搭建模型：
import torch.nn as nn
import torch
import torch.nn.functional as F
# 模型类
class GoogLeNet(nn.Module):
# 初始化（类别个数， 是否采用辅助分类器， 是否初始化权重）
def __init__(self, num_classes=1000, aux_logits=True, init_weights=False):
super(GoogLeNet, self).__init__()
self.aux_logits = aux_logits
# 这部分就按照论文中的结构，类似于搭积木那样搭建就可以了
self.conv1 = BasicConv2d(3, 64, kernel_size=7, stride=2, padding=3)
self.maxpool1 = nn.MaxPool2d(3, stride=2, ceil_mode=True) #ceil_mode=true代表向上取整
self.conv2 = BasicConv2d(64, 64, kernel_size=1)
self.conv3 = BasicConv2d(64, 192, kernel_size=3, padding=1)
self.maxpool2 = nn.MaxPool2d(3, stride=2, ceil_mode=True)
self.inception3a = Inception(192, 64, 96, 128, 16, 32, 32)
self.inception3b = Inception(256, 128, 128, 192, 32, 96, 64)
self.maxpool3 = nn.MaxPool2d(3, stride=2, ceil_mode=True)
self.inception4a = Inception(480, 192, 96, 208, 16, 48, 64)
self.inception4b = Inception(512, 160, 112, 224, 24, 64, 64)
self.inception4c = Inception(512, 128, 128, 256, 24, 64, 64)
self.inception4d = Inception(512, 112, 144, 288, 32, 64, 64)
self.inception4e = Inception(528, 256, 160, 320, 32, 128, 128)
self.maxpool4 = nn.MaxPool2d(3, stride=2, ceil_mode=True)
self.inception5a = Inception(832, 256, 160, 320, 32, 128, 128)
self.inception5b = Inception(832, 384, 192, 384, 48, 128, 128)
# 辅助分类器
if self.aux_logits:
self.aux1 = InceptionAux(512, num_classes)
self.aux2 = InceptionAux(528, num_classes)
self.avgpool = nn.AdaptiveAvgPool2d((1, 1))
self.dropout = nn.Dropout(0.4)
self.fc = nn.Linear(1024, num_classes)
# 初始化权重
if init_weights:
self._initialize_weights()
# 前向传播
def forward(self, x):
# N x 3 x 224 x 224
x = self.conv1(x)
# N x 64 x 112 x 112
x = self.maxpool1(x)
# N x 64 x 56 x 56
x = self.conv2(x)
# N x 64 x 56 x 56
x = self.conv3(x)
# N x 192 x 56 x 56
x = self.maxpool2(x)
# N x 192 x 28 x 28
x = self.inception3a(x)
# N x 256 x 28 x 28
x = self.inception3b(x)
# N x 480 x 28 x 28
x = self.maxpool3(x)
# N x 480 x 14 x 14
x = self.inception4a(x)
# N x 512 x 14 x 14
# 如果是训练模式，并且采用辅助分类器
if self.training and self.aux_logits: # 推理模式不需要辅助分类器
aux1 = self.aux1(x)
x = self.inception4b(x)
# N x 512 x 14 x 14
x = self.inception4c(x)
# N x 512 x 14 x 14
x = self.inception4d(x)
# N x 528 x 14 x 14
# 和上面一样
if self.training and self.aux_logits: # eval model lose this layer
aux2 = self.aux2(x)
x = self.inception4e(x)
# N x 832 x 14 x 14
x = self.maxpool4(x)
# N x 832 x 7 x 7
x = self.inception5a(x)
# N x 832 x 7 x 7
x = self.inception5b(x)
# N x 1024 x 7 x 7
x = self.avgpool(x)
# N x 1024 x 1 x 1
x = torch.flatten(x, 1)
# N x 1024
x = self.dropout(x)
x = self.fc(x)
# N x 1000 (num_classes)
# 在训练模式并且采用辅助分类器的情况下，返回主分类结果和两个辅助分类器的结果
if self.training and self.aux_logits: # eval model lose this layer
return x, aux2, aux1
return x
# 权重初始化函数
def _initialize_weights(self):
for m in self.modules(): # .modules()是继承于nn.moudle类的方法
# 返回的是由各个层组成的迭代器
if isinstance(m, nn.Conv2d): # 如果是卷积层
# 采用凯明初始化
nn.init.kaiming_normal_(m.weight, mode='fan_out', nonlinearity='relu')
if m.bias is not None:
# 将偏置初始化为0
nn.init.constant_(m.bias, 0)
elif isinstance(m, nn.Linear): # 如果是全连接层
# 权重均值是0， 标准差是0.01
nn.init.normal_(m.weight, 0, 0.01)
# 偏置是0
nn.init.constant_(m.bias, 0)
class Inception(nn.Module): # 定义inception结构
# 输入通道数，和七个输出通道数
def __init__(self, in_channels, ch1x1, ch3x3red, ch3x3, ch5x5red, ch5x5, pool_proj):
super(Inception, self).__init__()
self.branch1 = BasicConv2d(in_channels, ch1x1, kernel_size=1) #第一个分支
self.branch2 = nn.Sequential( #第二个分支
BasicConv2d(in_channels, ch3x3red, kernel_size=1),
BasicConv2d(ch3x3red, ch3x3, kernel_size=3, padding=1) # 保证输出大小等于输入大小
)
self.branch3 = nn.Sequential( # 第三个分支
BasicConv2d(in_channels, ch5x5red, kernel_size=1),
BasicConv2d(ch5x5red, ch5x5, kernel_size=5, padding=2) # 保证输出大小等于输入大小
)
self.branch4 = nn.Sequential( # 第四个分支
nn.MaxPool2d(kernel_size=3, stride=1, padding=1),
BasicConv2d(in_channels, pool_proj, kernel_size=1)
)
def forward(self, x):
branch1 = self.branch1(x) # 分别通过四个分支
branch2 = self.branch2(x)
branch3 = self.branch3(x)
branch4 = self.branch4(x)
outputs = [branch1, branch2, branch3, branch4]
return torch.cat(outputs, 1) #合并，在深度上进行拼接，深度会增加
class InceptionAux(nn.Module): #辅助分类器
def __init__(self, in_channels, num_classes):
super(InceptionAux, self).__init__()
self.averagePool = nn.AvgPool2d(kernel_size=5, stride=3) #平均池化下采样层
self.conv = BasicConv2d(in_channels, 128, kernel_size=1) # output[batch, 128, 4, 4]，卷积层
self.fc1 = nn.Linear(2048, 1024)
self.fc2 = nn.Linear(1024, num_classes)
def forward(self, x):
# aux1: N x 512 x 14 x 14, aux2: N x 528 x 14 x 14
x = self.averagePool(x)
# aux1: N x 512 x 4 x 4, aux2: N x 528 x 4 x 4
x = self.conv(x)
# N x 128 x 4 x 4
x = torch.flatten(x, 1) #将特征矩阵展平
x = F.dropout(x, 0.5, training=self.training) #随机失活50%的神经元，原论文中是70%
# 在训练集的时候，self.training=true，在验证集的时候，self.training=flase
# N x 2048
x = F.relu(self.fc1(x), inplace=True)
x = F.dropout(x, 0.5, training=self.training)
# N x 1024
x = self.fc2(x)
# N x num_classes
return x
class BasicConv2d(nn.Module): #定义一个类，将一个卷积层和激活函数组合在一起
# 输入通道数，输出通道数，卷积核大小，**是收集关键字参数，将他们组成一个字典
def __init__(self, in_channels, out_channels, **kwargs):
super(BasicConv2d, self).__init__()
self.conv = nn.Conv2d(in_channels, out_channels, **kwargs) # 卷积层
self.relu = nn.ReLU(inplace=True) # 激活函数，inplace = True可以节约内存
def forward(self, x):#定义一个正向传播过程
x = self.conv(x)
x = self.relu(x)
return x
训练：
import torch
import torch.nn as nn
from torchvision import transforms, datasets
import torchvision
import json
import matplotlib.pyplot as plt
import os
import torch.optim as optim
from model import GoogLeNet
device = torch.device("cuda:0" if torch.cuda.is_available() else "cpu")
print(device)
data_transform = {
"train": transforms.Compose([transforms.RandomResizedCrop(224), # 随机剪裁
transforms.RandomHorizontalFlip(), # 随机水平翻转
transforms.ToTensor(), # 转换为tensor
# ToTensor()的官方解释：Converts a PIL Image or numpy.ndarray (H x W x C) in the range
# [0, 255] to a torch.FloatTensor of shape (C x H x W) in the range [0.0, 1.0]
transforms.Normalize((0.5, 0.5, 0.5), (0.5, 0.5, 0.5))]), # 标准化，均值0，标准差1
"val": transforms.Compose([transforms.Resize((224, 224)),
transforms.ToTensor(),
transforms.Normalize((0.5, 0.5, 0.5), (0.5, 0.5, 0.5))])}
# data_root = os.path.abspath(os.path.join(os.getcwd(), "./")) # get data root path
# os.getcwd() 获得当前脚本所在文件夹的路径
# 在后面加上./
# 获取绝对路径
data_root = os.path.abspath("../")
image_path = data_root + "/data/flower_data/" # flower data set path，得到花文件的路径
# 在root指定的文件夹下寻找图片，对里面的图片做一定的变换
train_dataset = datasets.ImageFolder(root=image_path + "train",
transform=data_transform["train"])
# print(train_dataset.class_to_idx) {'daisy': 0, 'dandelion': 1, 'roses': 2, 'sunflowers': 3, 'tulips': 4}
# print(train_dataset.imgs) 会得到一个数组，数组中的每个元素是个元组，包括图片路径和标签
# [('E:\\pythonproject\\classical\\/data/flower_data/train\\daisy\\10140303196_b88d3d6cec.jpg', 0),
# ('E:\\pythonproject\\classical\\/data/flower_data/train\\daisy\\10172379554_b296050f82_n.jpg', 0), ...]
train_num = len(train_dataset)
# {'daisy':0, 'dandelion':1, 'roses':2, 'sunflower':3, 'tulips':4}
flower_list = train_dataset.class_to_idx
cla_dict = dict((val, key) for key, val in flower_list.items()) # 把花和数字的值反一下
# write dict into json file 搞一个json文件
json_str = json.dumps(cla_dict, indent=4)
with open('class_indices.json', 'w') as json_file:
json_file.write(json_str)
batch_size = 32
# DataLoader会把数据分成一批一批的，里面还有一些其他的设置参数
train_loader = torch.utils.data.DataLoader(train_dataset,
batch_size=batch_size, shuffle=True,
num_workers=0)
# 验证集的处理方式类似
validate_dataset = datasets.ImageFolder(root=image_path + "val",
transform=data_transform["val"])
val_num = len(validate_dataset)
validate_loader = torch.utils.data.DataLoader(validate_dataset,
batch_size=batch_size, shuffle=False,
num_workers=0)
# test_data_iter = iter(validate_loader)
# test_image, test_label = test_data_iter.next()
# net = torchvision.models.googlenet(num_classes=5)
# model_dict = net.state_dict()
# pretrain_model = torch.load("googlenet.pth")
# del_list = ["aux1.fc2.weight", "aux1.fc2.bias",
# "aux2.fc2.weight", "aux2.fc2.bias",
# "fc.weight", "fc.bias"]
# pretrain_dict = {k: v for k, v in pretrain_model.items() if k not in del_list}
# model_dict.update(pretrain_dict)
# net.load_state_dict(model_dict)
# 实例化模型
net = GoogLeNet(num_classes=5, aux_logits=True, init_weights=True)
net.to(device)
# 定义损失函数，定义优化器
loss_function = nn.CrossEntropyLoss()
optimizer = optim.Adam(net.parameters(), lr=0.0003)
# 最佳准确率
best_acc = 0.0
# 保存权重的路径
save_path = './googleNet.pth'
for epoch in range(30):
# train
net.train()
# 每个epoch清零损失
running_loss = 0.0
# 迭代器里面有索引和数据
for step, data in enumerate(train_loader, start=0):
# 数据是一个元组，有图像和标签
images, labels = data
# 每一个batch梯度清零
optimizer.zero_grad()
# 将图片输入网络，得到主分类器和两个辅助分类器的结果
logits, aux_logits2, aux_logits1 = net(images.to(device))
# 计算三个损失
# 这里的loss是每个batch中的平均loss
loss0 = loss_function(logits, labels.to(device))
loss1 = loss_function(aux_logits1, labels.to(device))
loss2 = loss_function(aux_logits2, labels.to(device))
# 三个损失加权求和
loss = loss0 + loss1 * 0.3 + loss2 * 0.3
# 反向传播求导
loss.backward()
# 沿梯度方向下降
optimizer.step()
# print statistics
# 累加每个batch的损失
running_loss += loss.item()
# print train process
rate = (step + 1) / len(train_loader)
a = "*" * int(rate * 50)
b = "." * int((1 - rate) * 50)
print("\rtrain loss: {:^3.0f}%[{}->{}]{:.3f}".format(int(rate * 100), a, b, loss), end="")
print()
# validate
net.eval()
acc = 0.0 # accumulate accurate number / epoch
with torch.no_grad(): # 不再计算梯度
for data_test in validate_loader:
test_images, test_labels = data_test
# 只有主分类器的结果
outputs = net(test_images.to(device)) # eval model only have last output layer
predict_y = torch.max(outputs, dim=1)[1]
# 计算所有预测正确的个数
acc += (predict_y == test_labels.to(device)).sum().item()
# 准确率
accurate_test = acc / val_num
# 如果准确率为历史最高值，就保存这次的训练结果
if accurate_test > best_acc:
torch.save(net.state_dict(), save_path)
best_acc = accurate_test # 更新最佳准确率
# 打印一次epoch之后的训练集平均损失和验证集的准确率
print('[epoch %d] train_loss: %.3f test_accuracy: %.3f' %
(epoch + 1, running_loss / step, accurate_test))
print('Finished Training')
预测：
import torch
from model import GoogLeNet
from PIL import Image
from torchvision import transforms
import matplotlib.pyplot as plt
import json
# 数据预处理
data_transform = transforms.Compose(
[transforms.Resize((224, 224)),
transforms.ToTensor(),
transforms.Normalize((0.5, 0.5, 0.5), (0.5, 0.5, 0.5))])
# load image
img = Image.open("../data/flower_predict/roses1.jpg")
plt.imshow(img)
# [N, C, H, W]
img = data_transform(img)
# expand batch dimension
img = torch.unsqueeze(img, dim=0)
# read class_indict
try:
json_file = open('./class_indices.json', 'r')
# 将json转化为字典
class_indict = json.load(json_file)
except Exception as e:
print(e)
exit(-1)
# create model
model = GoogLeNet(num_classes=5, aux_logits=False)
# load model weights
model_weight_path = "./googleNet.pth"
missing_keys, unexpected_keys = model.load_state_dict(torch.load(model_weight_path), strict=False)
model.eval()
with torch.no_grad():
# predict class
output = torch.squeeze(model(img))
predict = torch.softmax(output, dim=0)
predict_cla = torch.argmax(predict).numpy()
# 打印预测种类和概率值
print(class_indict[str(predict_cla)], predict[predict_cla].item())
plt.show()
扫一扫
1585
被折叠的 条评论
为什么被折叠?
到【灌水乐园】发言
