教程来自B站up@霹雳吧啦Wz,链接:
VGG网络详解及感受野的计算:https://www.bilibili.com/video/BV1q7411T7Y6
使用pytorch搭建VGG网络:https://www.bilibili.com/video/BV1i7411T7ZN
up的博客:https://blog.csdn.net/qq_37541097
论文链接:Very Deep Convolutional Networks for Large-Scale Image Recognition
VGGNet是牛津大学计算机视觉组VGG(Visual Geometry Group)和Google DeepMind公司的研究员仪器研发的深度卷积神经网络。VGG斩获2014年ImageNet竞赛中Localization Task(定位任务)第一名和Classification Task(分类任务)第二名。
配置深度从左(A)到右(E)依次增加,使用过程中常使用D配置(16层),此配置中包含13个卷积层和3个全连接层。
网络中的亮点:
通过堆叠多个3×3的卷积核来替代大尺度卷积核(减少所需参数),可以拥有相同的感受野
论文中提到,可以通过
堆叠两个3×3的卷积核替代5×5的卷积核
堆叠三个3×3的卷积核替代7×7的卷积核
在卷积神经网络中,决定某一层输出结果中一个元素所对应的输入层的区域大小,被称作感受野(receptive field)。通俗的解释是,输出feature map上的一个单元对应输入层上区域大小。
如下图所示,最下层为9×9×1的特征矩阵,通过卷积层Conv1(卷积核大小为3×3,步距为2),得到4×4×1,再经过池化层MaxPool1(池化核大小为2×2,步距为2),得到2×2×1的特征矩阵。
第三层中一个单元在第二层中对应的感受野是2×2区域,在原图(第一层)中对应的感受野大小为5×5区域。
感受野计算公式:
F ( i ) = ( F ( i + 1 ) − 1 ) × S t r i d e + K s i z e F(i) = (F(i+1)-1)×Stride+Ksize F(i)=(F(i+1)−1)×Stride+Ksize
Feature map: F ( 3 ) = 1 F(3)=1 F(3)=1
Pool1: F ( 2 ) = ( 1 − 1 ) × 2 + 2 = 2 F(2)=(1-1)×2+2=2 F(2)=(1−1)×2+2=2
Conv1: F ( 1 ) = ( 2 − 1 ) × 2 + 3 = 5 F(1)=(2-1)×2+3=5 F(1)=(2−1)×2+3=5
VGG网络中两个3×3的卷积核替代5×5的卷积核,三个3×3的卷积核替代7×7的卷积核,可以通过如下计算得出(VGG网络中步距默认为1)
假设一个特征矩阵通过三层3×3的卷积层后,得到feature map
Feature map: F ( 4 ) = 1 F(4)=1 F(4)=1
Conv3×3(3): F ( 3 ) = ( 1 − 1 ) × 1 + 3 = 3 F(3)=(1-1)×1+3=3 F(3)=(1−1)×1+3=3
Conv3×3(2): F ( 2 ) = ( 3 − 1 ) × 1 + 3 = 5 F(2)=(3-1)×1+3=5 F(2)=(3−1)×1+3=5
Conv3×3(1): F ( 1 ) = ( 5 − 1 ) × 1 + 3 = 7 F(1)=(5-1)×1+3=7 F(1)=(5−1)×1+3=7
通过三层3×3的卷积核卷积之后所得到的的一个单位对应的感受野相当于采用7×7的大小的卷积核得到的感受野相同。
使用7×7卷积核所需参数,与堆叠三个3×3卷积核所需的参数(假设输入输出channel为C)
7×7卷积核: 7 × 7 × C × C = 49 C 2 7×7×C×C=49C^2 7×7×C×C=49C2
三个3×3卷积核: 3 × 3 × C × C + 3 × 3 × C × C + 3 × 3 × C × C = 27 C 2 3×3×C×C+3×3×C×C+3×3×C×C=27C^2 3×3×C×C+3×3×C×C+3×3×C×C=27C2
所以两种方法的感受野相同,但堆叠3×3卷积核的方法所需参数更少。
输入为224×224大小的RGB图像,首先经过两个3×3的卷积层 → 一个最大下采样层 → 两个3×3的卷积层 → 一个最大下采样层 → 三个3×3的卷积层 → 一个最大下采样层 → 三个3×3的卷积层 → 一个最大下采样层 → 三个3×3的卷积层 → 一个最大下采样层 → 三个全连接层 → soft-max处理得到概率分布
表中 conv的Stride默认为1,padding默认为1
maxpool的Stride默认为2,size默认为2
由公式
o u t s i z e = ( i n s i z e − F s i z e + 2 P ) S + 1 out_{size}= \frac{(in_{size}-F_{size}+2P)}{S}+1 outsize=S(insize−Fsize+2P)+1
搭建 A、B、D、E 四个配置模型,将VGG网络分成两个部分:提取特征网络结构和分类网络结构。
model.py
import torch.nn as nn
import torch
# official pretrain weights
model_urls = {
'vgg11': 'https://download.pytorch.org/models/vgg11-bbd30ac9.pth',
'vgg13': 'https://download.pytorch.org/models/vgg13-c768596a.pth',
'vgg16': 'https://download.pytorch.org/models/vgg16-397923af.pth',
'vgg19': 'https://download.pytorch.org/models/vgg19-dcbb9e9d.pth'
}
class VGG(nn.Module):
def __init__(self, features, num_classes=1000, init_weights=False):
super(VGG, self).__init__()
self.features = features
# 构建分类网络结构
self.classifier = nn.Sequential(
nn.Linear(512*7*7, 4096), # 第一层全连接层
nn.ReLU(True),
nn.Dropout(p=0.5), # 50%的比例随机失活
nn.Linear(4096, 4096), # 第二层全连接层
nn.ReLU(True),
nn.Dropout(p=0.5),
nn.Linear(4096, num_classes) # 第三层全连接层
)
if init_weights: # 是否进行权重初始化
self._initialize_weights()
# 正向传播过程
def forward(self, x):
# N x 3 x 224 x 224
x = self.features(x) # 输入到特征提取网络
# N x 512 x 7 x 7
x = torch.flatten(x, start_dim=1) # 展平处理,从第1维度展平(第0维度为batch)
# N x 512*7*7
x = self.classifier(x) # 输入到分类网络中,得到输出
return x
def _initialize_weights(self):
for m in self.modules():
if isinstance(m, nn.Conv2d):
# nn.init.kaiming_normal_(m.weight, mode='fan_out', nonlinearity='relu')
nn.init.xavier_uniform_(m.weight)
if m.bias is not None:
nn.init.constant_(m.bias, 0)
elif isinstance(m, nn.Linear):
nn.init.xavier_uniform_(m.weight)
# nn.init.normal_(m.weight, 0, 0.01)
nn.init.constant_(m.bias, 0)
# 构建提取特征网络结构
def make_features(cfg: list): # 传入对应配置的列表
layers = [] # 定义空列表,存放每一层的结构
in_channels = 3 # 输入为RGB图片,输入通道为3
for v in cfg: # 遍历配置列表
if v == "M": # 如果为M,则为池化层,创建一个最大池化下采样层
layers += [nn.MaxPool2d(kernel_size=2, stride=2)]
else: # 不等于M,则为数字,创建卷积层
conv2d = nn.Conv2d(in_channels, v, kernel_size=3, padding=1)
layers += [conv2d, nn.ReLU(True)] # 每个卷积层都采用RELU激活函数,将定义好的卷积层和RELU拼接
in_channels = v
return nn.Sequential(*layers) # 非关键字参数,*layers可以传递任意数量的实参,以元组的形式导入
cfgs = {
'vgg11': [64, 'M', 128, 'M', 256, 256, 'M', 512, 512, 'M', 512, 512, 'M'],
'vgg13': [64, 64, 'M', 128, 128, 'M', 256, 256, 'M', 512, 512, 'M', 512, 512, 'M'],
'vgg16': [64, 64, 'M', 128, 128, 'M', 256, 256, 256, 'M', 512, 512, 512, 'M', 512, 512, 512, 'M'],
'vgg19': [64, 64, 'M', 128, 128, 'M', 256, 256, 256, 256, 'M', 512, 512, 512, 512, 'M', 512, 512, 512, 512, 'M'],
}
# 实例化配置模型
def vgg(model_name="vgg16", **kwargs):
assert model_name in cfgs, "Warning: model number {} not in cfgs dict!".format(model_name)
cfg = cfgs[model_name]
model = VGG(make_features(cfg), **kwargs) # 可以传递任意数量的实参,以字典的形式导入
return model
其中
cfgs为一个字典文件,字典中每个key代表一个模型的配置文件:‘vgg11’对应表格中的A,‘vgg13’对应表格中的B,‘vgg16’对应表格中的D,‘vgg19’对应表格中的E。每个key对应的值为一个列表,数字代表卷积层中卷积核的个数,字母‘M’表示池化层结构。
cfgs = {
'vgg11': [64, 'M', 128, 'M', 256, 256, 'M', 512, 512, 'M', 512, 512, 'M'],
'vgg13': [64, 64, 'M', 128, 128, 'M', 256, 256, 'M', 512, 512, 'M', 512, 512, 'M'],
'vgg16': [64, 64, 'M', 128, 128, 'M', 256, 256, 256, 'M', 512, 512, 512, 'M', 512, 512, 512, 'M'],
'vgg19': [64, 64, 'M', 128, 128, 'M', 256, 256, 256, 256, 'M', 512, 512, 512, 512, 'M', 512, 512, 512, 512, 'M'],
train.py与上一讲相同,只需改变部分参数。
import os
import json
import torch
import torch.nn as nn
from torchvision import transforms, datasets
import torch.optim as optim
from tqdm import tqdm
from model import vgg
def main():
device = torch.device("cuda:0" if torch.cuda.is_available() else "cpu")
print("using {} device.".format(device))
data_transform = {
"train": transforms.Compose([transforms.RandomResizedCrop(224),
transforms.RandomHorizontalFlip(),
transforms.ToTensor(),
transforms.Normalize((0.5, 0.5, 0.5), (0.5, 0.5, 0.5))]),
"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
image_path = os.path.join(data_root, "data_set", "flower_data") # flower data set path
assert os.path.exists(image_path), "{} path does not exist.".format(image_path)
train_dataset = datasets.ImageFolder(root=os.path.join(image_path, "train"),
transform=data_transform["train"])
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())
# write dict into json file
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
nw = min([os.cpu_count(), batch_size if batch_size > 1 else 0, 8]) # number of workers
print('Using {} dataloader workers every process'.format(nw))
train_loader = torch.utils.data.DataLoader(train_dataset,
batch_size=batch_size, shuffle=True,
num_workers=nw)
validate_dataset = datasets.ImageFolder(root=os.path.join(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=nw)
print("using {} images for training, {} images for validation.".format(train_num,
val_num))
# test_data_iter = iter(validate_loader)
# test_image, test_label = test_data_iter.next()
model_name = "vgg16"
net = vgg(model_name=model_name, num_classes=5, init_weights=True)
net.to(device)
loss_function = nn.CrossEntropyLoss()
optimizer = optim.Adam(net.parameters(), lr=0.0001)
epochs = 30
best_acc = 0.0
save_path = './{}Net.pth'.format(model_name)
train_steps = len(train_loader)
for epoch in range(epochs):
# train
net.train()
running_loss = 0.0
train_bar = tqdm(train_loader)
for step, data in enumerate(train_bar):
images, labels = data
optimizer.zero_grad()
outputs = net(images.to(device))
loss = loss_function(outputs, labels.to(device))
loss.backward()
optimizer.step()
# print statistics
running_loss += loss.item()
train_bar.desc = "train epoch[{}/{}] loss:{:.3f}".format(epoch + 1,
epochs,
loss)
# validate
net.eval()
acc = 0.0 # accumulate accurate number / epoch
with torch.no_grad():
val_bar = tqdm(validate_loader)
for val_data in val_bar:
val_images, val_labels = val_data
outputs = net(val_images.to(device))
predict_y = torch.max(outputs, dim=1)[1]
acc += torch.eq(predict_y, val_labels.to(device)).sum().item()
val_accurate = acc / val_num
print('[epoch %d] train_loss: %.3f val_accuracy: %.3f' %
(epoch + 1, running_loss / train_steps, val_accurate))
if val_accurate > best_acc:
best_acc = val_accurate
torch.save(net.state_dict(), save_path)
print('Finished Training')
if __name__ == '__main__':
main()
predict.py
import os
import json
import torch
from PIL import Image
from torchvision import transforms
import matplotlib.pyplot as plt
from model import vgg
def main():
device = torch.device("cuda:0" if torch.cuda.is_available() else "cpu")
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_path = "../tulip.jpg"
assert os.path.exists(img_path), "file: '{}' dose not exist.".format(img_path)
img = Image.open(img_path)
plt.imshow(img)
# [N, C, H, W]
img = data_transform(img)
# expand batch dimension
img = torch.unsqueeze(img, dim=0)
# read class_indict
json_path = './class_indices.json'
assert os.path.exists(json_path), "file: '{}' dose not exist.".format(json_path)
json_file = open(json_path, "r")
class_indict = json.load(json_file)
# create model
model = vgg(model_name="vgg16", num_classes=5).to(device)
# load model weights
weights_path = "./vgg16Net.pth"
assert os.path.exists(weights_path), "file: '{}' dose not exist.".format(weights_path)
model.load_state_dict(torch.load(weights_path, map_location=device))
model.eval()
with torch.no_grad():
# predict class
output = torch.squeeze(model(img.to(device))).cpu()
predict = torch.softmax(output, dim=0)
predict_cla = torch.argmax(predict).numpy()
print_res = "class: {} prob: {:.3}".format(class_indict[str(predict_cla)],
predict[predict_cla].numpy())
plt.title(print_res)
print(print_res)
plt.show()
if __name__ == '__main__':
main()