基于ResNet网络架构训练图像分类模型
目录
本项目通过训练数据6552条，验证数据818条，通过迁移学习ResNet经典网络架构，实现对102种花的照片进行分类。流程：利用torchvision库中transforms模块进行数据的增强和预处理；然后调用torchvision库中的ResNet经典网络架构，用人家训练好的权重参数来提取特征（迁移别人的卷积层）；重新加入全连接层传入自己的分类数；然后，训练自己的全连接层；接着，训练所有层；最后，测试网络效果。
1.数据预处理
(1)导包
import os
import matplotlib.pyplot as plt
%matplotlib inline
import numpy as np
import torch
from torch import nn
import torch.optim as optim
import torchvision
#pip install torchvision
from torchvision import transforms, models, datasets
#https://pytorch.org/docs/stable/torchvision/index.html
import imageio
import time
import warnings
import random
import sys
import copy
import json
from PIL import Image
(2)数据读取
指定数据对应的文件夹位置
data_dir = './flower_data/'
train_dir = data_dir + '/train'
valid_dir = data_dir + '/valid'
(3)制作数据
data_transforms = {#指定了所有图像预处理操作
'train': transforms.Compose([transforms.RandomRotation(45),#随机旋转，-45到45度之间随机选
transforms.CenterCrop(224),#从中心开始裁剪
transforms.RandomHorizontalFlip(p=0.5),#随机水平翻转 选择一个概率概率，p=0.5百分之50的概率会执行水平翻转
transforms.RandomVerticalFlip(p=0.5),#随机垂直翻转
transforms.ColorJitter(brightness=0.2, contrast=0.1, saturation=0.1, hue=0.1),#参数1为亮度，参数2为对比度，参数3为饱和度，参数4为色相
transforms.RandomGrayscale(p=0.025),#概率转换成灰度率，3通道就是R=G=B
transforms.ToTensor(),#转换成tensor格式
transforms.Normalize([0.485, 0.456, 0.406], [0.229, 0.224, 0.225])#均值，标准差
]),
'valid': transforms.Compose([transforms.Resize(256),
transforms.CenterCrop(224),
transforms.ToTensor(),
transforms.Normalize([0.485, 0.456, 0.406], [0.229, 0.224, 0.225])
]),
}
查看数据：训练集有6552条数据，验证集有818条数据。
加载数据集
batch_size = 8
image_datasets = {x: datasets.ImageFolder(os.path.join(data_dir, x), data_transforms[x]) for x in ['train', 'valid']}#用datasets.ImageFolder读数据，传入路径和预处理方法，构建完数据集
dataloaders = {x: torch.utils.data.DataLoader(image_datasets[x], batch_size=batch_size, shuffle=True) for x in ['train', 'valid']}#一批一批取数据
dataset_sizes = {x: len(image_datasets[x]) for x in ['train', 'valid']}
class_names = image_datasets['train'].classes
(4)读取标签对应的实际名字
cat_to_name.json文件中保存了每一个序号对应的花的名字。
with open('cat_to_name.json', 'r') as f:
cat_to_name = json.load(f)
部分文件内容：
(5)展示数据
数据展示需要将tensor的数据需要转换成numpy的格式，而且还需要还原回标准化的结果
def im_convert(tensor):
""" 展示数据"""
image = tensor.to("cpu").clone().detach()
image = image.numpy().squeeze()
image = image.transpose(1,2,0)#还原颜色通道顺序，从chw->hwc
image = image * np.array((0.229, 0.224, 0.225)) + np.array((0.485, 0.456, 0.406))#还原，乘均值，加标准差
image = image.clip(0, 1)
return image
#绘图操作
fig=plt.figure(figsize=(20, 12))
columns = 4
rows = 2
dataiter = iter(dataloaders['valid'])#迭代一次，取一批数据
inputs, classes = dataiter.next()#返回输入图像，标签值
for idx in range (columns*rows):
ax = fig.add_subplot(rows, columns, idx+1, xticks=[], yticks=[])#画布划分
ax.set_title(cat_to_name[str(int(class_names[classes[idx]]))])
plt.imshow(im_convert(inputs[idx]))
plt.show()
2.构建网络
(1)加载已有网络模型
model_name = 'resnet' #可选的比较多 ['resnet', 'alexnet', 'vgg', 'squeezenet', 'densenet', 'inception']
#是否用人家训练好的特征来做
feature_extract = True #用别人训练好的权重
# 是否用GPU训练
train_on_gpu = torch.cuda.is_available()
if not train_on_gpu:
print('CUDA is not available. Training on CPU ...')
else:
print('CUDA is available! Training on GPU ...')
device = torch.device("cuda:0" if torch.cuda.is_available() else "cpu")
model_ft = models.resnet152()#加载网络模型resnet152层
截取网络最后的部分，可以看到加载的resnet152网络模型最后是1000分类。
(2)选择需不需要冻住哪些层
def set_parameter_requires_grad(model, feature_extracting):#迁移学习，迁移别人模型时，选择冻住哪些层
if feature_extracting:
for param in model.parameters():
param.requires_grad = False #指定哪些层不做训练，不做梯度更新
(3)重新加入全连接层传入自己的分类数
##构建网络（1.加载已有网络模型，2.选择需不需要冻住哪些层，3.重新加入全连接层传入自己的分类数）
def initialize_model(model_name, num_classes, feature_extract, use_pretrained=True):#num_classes：类别个数
# 选择合适的模型，不同模型的初始化方法稍微有点区别
model_ft = None
input_size = 0
if model_name == "resnet":
""" Resnet152
"""
model_ft = models.resnet152(pretrained=use_pretrained)#加载训练好的模型
set_parameter_requires_grad(model_ft, feature_extract)#迁移学习，选择性地冻住一些层
num_ftrs = model_ft.fc.in_features #拿到最后一层的fc.in_features =2048
model_ft.fc = nn.Sequential(nn.Linear(num_ftrs, 102),#重新加入最后的全连接层，传入特征数num_ftrs：2048和分类数102
nn.LogSoftmax(dim=1)) #后续直接用LogSoftmax算损失，不用交叉熵
input_size = 224
elif model_name == "alexnet":
""" Alexnet
"""
model_ft = models.alexnet(pretrained=use_pretrained)
set_parameter_requires_grad(model_ft, feature_extract)
num_ftrs = model_ft.classifier[6].in_features
model_ft.classifier[6] = nn.Linear(num_ftrs,num_classes)
input_size = 224
elif model_name == "vgg":
""" VGG11_bn
"""
model_ft = models.vgg16(pretrained=use_pretrained)
set_parameter_requires_grad(model_ft, feature_extract)
num_ftrs = model_ft.classifier[6].in_features
model_ft.classifier[6] = nn.Linear(num_ftrs,num_classes)
input_size = 224
elif model_name == "squeezenet":
""" Squeezenet
"""
model_ft = models.squeezenet1_0(pretrained=use_pretrained)
set_parameter_requires_grad(model_ft, feature_extract)
model_ft.classifier[1] = nn.Conv2d(512, num_classes, kernel_size=(1,1), stride=(1,1))
model_ft.num_classes = num_classes
input_size = 224
elif model_name == "densenet":
""" Densenet
"""
model_ft = models.densenet121(pretrained=use_pretrained)
set_parameter_requires_grad(model_ft, feature_extract)
num_ftrs = model_ft.classifier.in_features
model_ft.classifier = nn.Linear(num_ftrs, num_classes)
input_size = 224
elif model_name == "inception":
""" Inception v3
Be careful, expects (299,299) sized images and has auxiliary output
"""
model_ft = models.inception_v3(pretrained=use_pretrained)
set_parameter_requires_grad(model_ft, feature_extract)
# Handle the auxilary net
num_ftrs = model_ft.AuxLogits.fc.in_features
model_ft.AuxLogits.fc = nn.Linear(num_ftrs, num_classes)
# Handle the primary net
num_ftrs = model_ft.fc.in_features
model_ft.fc = nn.Linear(num_ftrs,num_classes)
input_size = 299
else:
print("Invalid model name, exiting...")
exit()
return model_ft, input_size
(4)网络搭建结果
model_ft, input_size = initialize_model(model_name, 102, feature_extract, use_pretrained=True)#feature_extract：要不要冻一些层，use_pretrained=True：用别人训练好的模型
#GPU计算
model_ft = model_ft.to(device)
# 模型保存
filename='checkpoint.pth'#保存训练结果，后续测试可直接用该模型
# 是否训练所有层
params_to_update = model_ft.parameters()
print("Params to learn:")
if feature_extract:
params_to_update = []
for name,param in model_ft.named_parameters():
if param.requires_grad == True:
params_to_update.append(param)
print("\t",name)
else:
for name,param in model_ft.named_parameters():
if param.requires_grad == True:
print("\t",name)#本项目用别人训练好的权重参数，只有全连接层的权重参数和偏置参数需要学习
搭建完自己的全连接层结果展示：可见全连接层的分类数已经改成对应自己项目的102分类
3.训练自己的全连接层
# 优化器设置
optimizer_ft = optim.Adam(params_to_update, lr=1e-2)#lr学习率
scheduler = optim.lr_scheduler.StepLR(optimizer_ft, step_size=7, gamma=0.1)#学习率每7个epoch衰减成原来的1/10，加入学习率衰减更稳定
#最后一层已经LogSoftmax()了，所以不能nn.CrossEntropyLoss()来计算了，nn.CrossEntropyLoss()相当于logSoftmax()和nn.NLLLoss()整合
criterion = nn.NLLLoss()#损失函数
def train_model(model, dataloaders, criterion, optimizer, num_epochs=25, is_inception=False,filename=filename):
#传入参数：模型，一批一批取数据，损失函数，优化器，训练次数，是否用其他网络，
since = time.time()
best_acc = 0#最好准确率
"""
checkpoint = torch.load(filename)
best_acc = checkpoint['best_acc']
model.load_state_dict(checkpoint['state_dict'])
optimizer.load_state_dict(checkpoint['optimizer'])
model.class_to_idx = checkpoint['mapping']
"""
model.to(device)
val_acc_history = []
train_acc_history = []
train_losses = []
valid_losses = []
LRs = [optimizer.param_groups[0]['lr']]#学习率
best_model_wts = copy.deepcopy(model.state_dict())#实时更新，单独存一下学习得最好的权重参数
for epoch in range(num_epochs):
print('Epoch {}/{}'.format(epoch, num_epochs - 1))
print('-' * 10)
# 训练和验证
for phase in ['train', 'valid']:
if phase == 'train':
model.train() # 训练
else:
model.eval() # 验证
running_loss = 0.0
running_corrects = 0
# 把数据都取个遍
for inputs, labels in dataloaders[phase]:
inputs = inputs.to(device)#to(device)表示传GPU
labels = labels.to(device)
# 清零
optimizer.zero_grad()
# 只有训练的时候计算和更新梯度
with torch.set_grad_enabled(phase == 'train'):
if is_inception and phase == 'train':#执行不到
outputs, aux_outputs = model(inputs)
loss1 = criterion(outputs, labels)
loss2 = criterion(aux_outputs, labels)
loss = loss1 + 0.4*loss2
else:#resnet执行的是这里
outputs = model(inputs)
loss = criterion(outputs, labels)#计算损失
_, preds = torch.max(outputs, 1)#拿到当前预测值中概率最大的
# 训练阶段更新权重
if phase == 'train':
loss.backward()
optimizer.step()
# 计算损失
running_loss += loss.item() * inputs.size(0)
running_corrects += torch.sum(preds == labels.data)#统计预测对的个数
#打印操作，准确率
epoch_loss = running_loss / len(dataloaders[phase].dataset)
epoch_acc = running_corrects.double() / len(dataloaders[phase].dataset)
time_elapsed = time.time() - since
print('Time elapsed {:.0f}m {:.0f}s'.format(time_elapsed // 60, time_elapsed % 60))
print('{} Loss: {:.4f} Acc: {:.4f}'.format(phase, epoch_loss, epoch_acc))
# 得到最好那次的模型
if phase == 'valid' and epoch_acc > best_acc:#在验证集上，若当前准确率高于记录的最后准确率：
best_acc = epoch_acc
best_model_wts = copy.deepcopy(model.state_dict())#copy当前模型的权重参数
state = {#保存指标
'state_dict': model.state_dict(),#模型所有参数
'best_acc': best_acc,#准确率
'optimizer' : optimizer.state_dict(),#优化器
}
torch.save(state, filename)#保存文件
if phase == 'valid':
val_acc_history.append(epoch_acc)
valid_losses.append(epoch_loss)
scheduler.step(epoch_loss)
if phase == 'train':
train_acc_history.append(epoch_acc)
train_losses.append(epoch_loss)
#打印
print('Optimizer learning rate : {:.7f}'.format(optimizer.param_groups[0]['lr']))
LRs.append(optimizer.param_groups[0]['lr'])
print()
time_elapsed = time.time() - since
print('Training complete in {:.0f}m {:.0f}s'.format(time_elapsed // 60, time_elapsed % 60))
print('Best val Acc: {:4f}'.format(best_acc))
# 训练完后用最好的一次当做模型最终的结果
model.load_state_dict(best_model_wts)
return model, val_acc_history, train_acc_history, valid_losses, train_losses, LRs
#训练
model_ft, val_acc_history, train_acc_history, valid_losses, train_losses, LRs = train_model(model_ft, dataloaders, criterion, optimizer_ft, num_epochs=20, is_inception=(model_name=="inception"))
#num_epochs=20 迭代次数，根据电脑性能可改5-50
训练结果：
4.训练所有层
for param in model_ft.parameters():
param.requires_grad = True#所有参数都训练
# 再继续训练所有的参数，学习率调小一点
optimizer = optim.Adam(params_to_update, lr=1e-4)
scheduler = optim.lr_scheduler.StepLR(optimizer_ft, step_size=7, gamma=0.1)
# 损失函数
criterion = nn.NLLLoss()
# Load the checkpoint 在之前基础上训练，之前保存了最好的模型结果
checkpoint = torch.load(filename)#传入路径
best_acc = checkpoint['best_acc']#最好准确率
model_ft.load_state_dict(checkpoint['state_dict'])#读入最好的模型参数
optimizer.load_state_dict(checkpoint['optimizer'])
#model_ft.class_to_idx = checkpoint['mapping']
model_ft, val_acc_history, train_acc_history, valid_losses, train_losses, LRs = train_model(model_ft, dataloaders, criterion, optimizer, num_epochs=10, is_inception=(model_name=="inception"))
训练结果：
5.测试网络效果
(1)加载训练好的模型
model_ft, input_size = initialize_model(model_name, 102, feature_extract, use_pretrained=True)#加载模型
# GPU模式
model_ft = model_ft.to(device)
# 保存文件的名字
filename='seriouscheckpoint.pth'
# 加载模型
checkpoint = torch.load(filename)
best_acc = checkpoint['best_acc']
model_ft.load_state_dict(checkpoint['state_dict'])
(2)测试数据预处理
- 测试数据处理方法需要跟训练时一样才可以
- crop操作的目的是保证输入的大小是一致的
- 标准化操作也是必须的，用跟训练数据相同的mean和std,但是需要注意一点训练数据是在0-1上进行标准化，所以测试数据也需要先归一化
- 最后一点，PyTorch中颜色通道是第一个维度，跟很多工具包都不一样，需要转换
def process_image(image_path):#预处理图像
# 读取测试数据
img = Image.open(image_path)
# Resize,thumbnail方法只能进行缩小，所以进行了判断
if img.size[0] > img.size[1]:
img.thumbnail((10000, 256))
else:
img.thumbnail((256, 10000))
# Crop操作，把图像裁剪成224*224
left_margin = (img.width-224)/2
bottom_margin = (img.height-224)/2
right_margin = left_margin + 224
top_margin = bottom_margin + 224
img = img.crop((left_margin, bottom_margin, right_margin,
top_margin))
# 相同的预处理方法
img = np.array(img)/255#归一化，压缩到0-1之间
mean = np.array([0.485, 0.456, 0.406]) #provided mean与训练集保持一致
std = np.array([0.229, 0.224, 0.225]) #provided std
img = (img - mean)/std
# 注意颜色通道应该放在第一个位置cwh
img = img.transpose((2, 0, 1))
return img
def imshow(image, ax=None, title=None):
"""展示数据"""
if ax is None:
fig, ax = plt.subplots()
# 颜色通道还原
image = np.array(image).transpose((1, 2, 0))
# 预处理还原
mean = np.array([0.485, 0.456, 0.406])
std = np.array([0.229, 0.224, 0.225])
image = std * image + mean
image = np.clip(image, 0, 1)
ax.imshow(image)
ax.set_title(title)
return ax
# 得到一个batch的测试数据
dataiter = iter(dataloaders['valid'])
images, labels = dataiter.next()
model_ft.eval()
if train_on_gpu:
output = model_ft(images.cuda())
else:
output = model_ft(images)
output表示对一个batch中每一个数据得到其属于各个类别的可能性
(3)得到概率最大的那个
_, preds_tensor = torch.max(output, 1)
preds = np.squeeze(preds_tensor.numpy()) if not train_on_gpu else np.squeeze(preds_tensor.cpu().numpy())
(4)展示预测结果
fig=plt.figure(figsize=(20, 20))
columns =4
rows = 2
for idx in range (columns*rows):
ax = fig.add_subplot(rows, columns, idx+1, xticks=[], yticks=[])
plt.imshow(im_convert(images[idx]))
ax.set_title("{} ({})".format(cat_to_name[str(preds[idx])], cat_to_name[str(labels[idx].item())]),
color=("green" if cat_to_name[str(preds[idx])]==cat_to_name[str(labels[idx].item())] else "red"))#进行判断，绿色表示预测正确，红色表示预测错误。
plt.show()
结果：
