>- **🍨 本文为(🔗365天深度学习训练营)中的学习记录博客**
>- **🍦 参考文章:[Pytorch实战 | 第P5周:运动鞋识别]
>- **🍖 原作者:[K同学啊 | 接辅导、项目定制]
>- **🚀 文章来源:[K同学的学习圈子]
🍺 要求:
- 自己搭建VGG-16网络框架(完成)
- 调用官方的VGG-16网络框架(P6已完成)
- 如何查看模型的参数量以及相关指标 (完成)
🍻 拔高(可选):
- 验证集准确率达到100% (达到了99.2%)
- 使用PPT画出VGG-16算法框架图(发论文需要这项技能)(完成)
🔎 探索(难度有点大)
- 在不影响准确率的前提下轻量化模型
- 目前VGG16的Total params是134,276,932
一、前期准备
1.设置GPU
import torch
import torchvision
if __name__=='__main__':
device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
print("Using {} device\n".format(device))
Using cuda device
2.导入本地数据并划分数据集
import pathlib
def Localdata(data_dir):
data_dir = pathlib.Path(data_dir)
data_path = list(data_dir.glob("*"))
ClassNames = [str(path).split("\\")[-1] for path in data_path]
train_transforms = torchvision.transforms.Compose([
torchvision.transforms.Resize([224,224]),
torchvision.transforms.ToTensor(),#把PIL Image或numpy.ndaaray转换成tensor
torchvision.transforms.Normalize(
mean=[0.486, 0.456, 0.406],
std=[0.229, 0.224, 0.225])
])
total_dataset = torchvision.datasets.ImageFolder(data_dir,transform=train_transforms)
print(total_dataset,'\n')
print(total_dataset.class_to_idx,'\n')
train_size = int(0.8* len(total_dataset))
test_size = len(total_dataset)-train_size
print('train_size:',train_size,'test_size:',test_size,'\n')
train_dataset,test_dataset = torch.utils.data.random_split(total_dataset,[train_size,test_size])
return ClassNames,train_dataset,test_dataset
import os
if __name__=='__main__':
root = 'G://'
output = 'output'
data_dir = os.path.join(root,'49-data')
batch_size = 32
ClassNames,train_ds,test_ds = Localdata(data_dir)
'''图片类别数'''
num_classes = len(ClassNames)
print("num_classes:{0}\n".format(num_classes))
Dataset ImageFolder Number of datapoints: 1200 Root location: G:\49-data StandardTransform Transform: Compose( Resize(size=[224, 224], interpolation=bilinear, max_size=None, antialias=warn) ToTensor() Normalize(mean=[0.486, 0.456, 0.406], std=[0.229, 0.224, 0.225]) ) {'Dark': 0, 'Green': 1, 'Light': 2, 'Medium': 3} train_size: 960 test_size: 240 num_classes:4
3.数据加载+数据查看
'''加载数据,并设置batch_size'''
import matplotlib.pyplot as plt
def loadData(train_ds,test_ds,batch_size=32,root='',show_flag=False):
train_dl = torch.utils.data.DataLoader(train_ds,
batch_size=batch_size,
shuffle=True,
num_workers=2)
test_dl = torch.utils.data.DataLoader(test_ds,
batch_size=batch_size,
shuffle=True,
num_workers=2)
for X,y in test_dl:
print("shape of X[N,C,H,W]:",X.shape)
print("shape of y:",y.shape,y.dtype,'\n')
break
imgs, labels = next(iter(train_dl))
print('Image shape:',imgs.shape,'\n')
displayData(imgs,root,show_flag)
return train_dl,test_dl
'''数据可视化'''
def displayData(imgs,root='',flag=False):
plt.figure('Data Visualization',figsize=(20,5))
for i, imgs in enumerate(imgs[:20]):
#调整维度顺序,将[3,224,224]调整为[224,224,3]
npimg = imgs.numpy().transpose((1,2,0))
plt.subplot(2,10,i+1)
plt.imshow(npimg)
plt.axis('off')
plt.savefig(os.path.join(root,'DatasetDisplay.png'))
if flag:
plt.show()
else:
plt.close("all")
train_dl,test_dl = loadData(train_ds,test_ds,batch_size,root,True)
shape of X[N,C,H,W]: torch.Size([32, 3, 224, 224]) shape of y: torch.Size([32]) torch.int64
二、手动搭建 VGG-16模型
关于VGG-16模型,网上已经有很多资料对其进行详述了,具体可以参考网站:VGGNet-16 Architecture: A Complete Guide | Kaggle
以上图片为VGG-16算法框架
import torch.nn as nn
import torchsummary
class VGG16(nn.Module):
def __init__(self):
super(VGG16,self).__init__()
#卷积块1
self.block1 = nn.Sequential(
nn.Conv2d(3, 64, kernel_size=(3,3), stride=(1,1), padding=(1,1)),
nn.ReLU(),
nn.Conv2d(64, 64, kernel_size=(3,3), stride=(1,1), padding=(1,1)),
nn.ReLU(),
nn.MaxPool2d(kernel_size=(2,2), stride=(2,2))
)
#卷积块2
self.block2 = nn.Sequential(
nn.Conv2d(64, 128, kernel_size=(3,3), stride=(1,1), padding=(1,1)),
nn.ReLU(),
nn.Conv2d(128, 128, kernel_size=(3,3), stride=(1,1), padding=(1,1)),
nn.ReLU(),
nn.MaxPool2d(kernel_size=(2,2),stride=(2,2))
)
#卷积块3
self.block3 = nn.Sequential(
nn.Conv2d(128, 256, kernel_size=(3,3), stride=(1,1), padding=(1,1)),
nn.ReLU(),
nn.Conv2d(256, 256, kernel_size=(3,3), stride=(1,1), padding=(1,1)),
nn.ReLU(),
nn.Conv2d(256, 256, kernel_size=(3,3), stride=(1,1), padding=(1,1)),
nn.ReLU(),
nn.MaxPool2d(kernel_size=(2,2),stride=(2,2))
)
#卷积块4
self.block4 = nn.Sequential(nn.Conv2d(256, 512, kernel_size=(3,3), stride=(1,1), padding=(1,1)),
nn.ReLU(),
nn.Conv2d(512, 512, kernel_size=(3,3), stride=(1,1), padding=(1,1)),
nn.ReLU(),
nn.Conv2d(512, 512, kernel_size=(3,3), stride=(1,1), padding=(1,1)),
nn.ReLU(),
nn.MaxPool2d(kernel_size=(2,2),stride=(2,2))
)
#卷积块5
self.block5 = nn.Sequential(
nn.Conv2d(512, 512, kernel_size=(3,3),stride=(1,1), padding=(1,1)),
nn.ReLU(),
nn.Conv2d(512, 512, kernel_size=(3,3), stride=(1,1), padding=(1,1)),
nn.ReLU(),
nn.Conv2d(512, 512, kernel_size=(3,3), stride=(1,1), padding=(1,1)),
nn.ReLU(),
nn.MaxPool2d(kernel_size=(2,2), stride=(2,2))
)
#全连接网络层,用于分类
self.classifier = nn.Sequential(
nn.Linear(in_features=512*7*7,out_features=4096),
nn.ReLU(),
nn.Linear(in_features=4096, out_features=4096),
nn.ReLU(),
nn.Linear(in_features=4096, out_features=num_classes)
)
def forward(self, x):
x = self.block1(x)
x = self.block2(x)
x = self.block3(x)
x = self.block4(x)
x = self.block5(x)
x = torch.flatten(x, start_dim=1)
x = self.classifier(x)
return x
if __name__ == '__main__':
'''设置GPU'''
device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
print("Using {} device\n".format(device))
model = VGG16().to(device)
torchsummary.summary(model,(3, 224, 224))
print(model)
Using cuda device ---------------------------------------------------------------- Layer (type) Output Shape Param # ================================================================ Conv2d-1 [-1, 64, 224, 224] 1,792 ReLU-2 [-1, 64, 224, 224] 0 Conv2d-3 [-1, 64, 224, 224] 36,928 ReLU-4 [-1, 64, 224, 224] 0 MaxPool2d-5 [-1, 64, 112, 112] 0 Conv2d-6 [-1, 128, 112, 112] 73,856 ReLU-7 [-1, 128, 112, 112] 0 Conv2d-8 [-1, 128, 112, 112] 147,584 ReLU-9 [-1, 128, 112, 112] 0 MaxPool2d-10 [-1, 128, 56, 56] 0 Conv2d-11 [-1, 256, 56, 56] 295,168 ReLU-12 [-1, 256, 56, 56] 0 Conv2d-13 [-1, 256, 56, 56] 590,080 ReLU-14 [-1, 256, 56, 56] 0 Conv2d-15 [-1, 256, 56, 56] 590,080 ReLU-16 [-1, 256, 56, 56] 0 MaxPool2d-17 [-1, 256, 28, 28] 0 Conv2d-18 [-1, 512, 28, 28] 1,180,160 ReLU-19 [-1, 512, 28, 28] 0 Conv2d-20 [-1, 512, 28, 28] 2,359,808 ReLU-21 [-1, 512, 28, 28] 0 Conv2d-22 [-1, 512, 28, 28] 2,359,808 ReLU-23 [-1, 512, 28, 28] 0 MaxPool2d-24 [-1, 512, 14, 14] 0 Conv2d-25 [-1, 512, 14, 14] 2,359,808 ReLU-26 [-1, 512, 14, 14] 0 Conv2d-27 [-1, 512, 14, 14] 2,359,808 ReLU-28 [-1, 512, 14, 14] 0 Conv2d-29 [-1, 512, 14, 14] 2,359,808 ReLU-30 [-1, 512, 14, 14] 0 MaxPool2d-31 [-1, 512, 7, 7] 0 Linear-32 [-1, 4096] 102,764,544 ReLU-33 [-1, 4096] 0 Linear-34 [-1, 4096] 16,781,312 ReLU-35 [-1, 4096] 0 Linear-36 [-1, 4] 16,388 ================================================================ Total params: 134,276,932 Trainable params: 134,276,932 Non-trainable params: 0 ---------------------------------------------------------------- Input size (MB): 0.57 Forward/backward pass size (MB): 218.52 Params size (MB): 512.23 Estimated Total Size (MB): 731.32 ---------------------------------------------------------------- VGG16( (block1): Sequential( (0): Conv2d(3, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1)) (1): ReLU() (2): Conv2d(64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1)) (3): ReLU() (4): MaxPool2d(kernel_size=(2, 2), stride=(2, 2), padding=0, dilation=1, ceil_mode=False) ) (block2): Sequential( (0): Conv2d(64, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1)) (1): ReLU() (2): Conv2d(128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1)) (3): ReLU() (4): MaxPool2d(kernel_size=(2, 2), stride=(2, 2), padding=0, dilation=1, ceil_mode=False) ) (block3): Sequential( (0): Conv2d(128, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1)) (1): ReLU() (2): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1)) (3): ReLU() (4): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1)) (5): ReLU() (6): MaxPool2d(kernel_size=(2, 2), stride=(2, 2), padding=0, dilation=1, ceil_mode=False) ) (block4): Sequential( (0): Conv2d(256, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1)) (1): ReLU() (2): Conv2d(512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1)) (3): ReLU() (4): Conv2d(512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1)) (5): ReLU() (6): MaxPool2d(kernel_size=(2, 2), stride=(2, 2), padding=0, dilation=1, ceil_mode=False) ) (block5): Sequential( (0): Conv2d(512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1)) (1): ReLU() (2): Conv2d(512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1)) (3): ReLU() (4): Conv2d(512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1)) (5): ReLU() (6): MaxPool2d(kernel_size=(2, 2), stride=(2, 2), padding=0, dilation=1, ceil_mode=False) ) (classifier): Sequential( (0): Linear(in_features=25088, out_features=4096, bias=True) (1): ReLU() (2): Linear(in_features=4096, out_features=4096, bias=True) (3): ReLU() (4): Linear(in_features=4096, out_features=4, bias=True) ) )
三、训练模型
1.编写训练函数
optimizer.zero_grad()
loss.backward()
optimizer.step()这三个函数是做参数更新,分别表示把1.梯度清零 2.反向传播 3.梯度更新
#训练循环
def train(dataloader,model,optimizer,loss_fn):
size = len(dataloader.dataset)
num_batches = len(dataloader)
train_loss, train_acc = 0, 0
for X,y in dataloader:
X, y = X.to(device),y.to(device)
#计算误差
pred = model(X)
loss = loss_fn(pred, y)
#反向传播
optimizer.zero_grad()
loss.backward()
optimizer.step()
#记录acc与loss
train_acc += (pred.argmax(1) == y).type(torch.float).sum().item()
train_loss += loss.item()
train_acc /= size
train_loss /= num_batches
return train_acc,train_loss
2.编写测试函数
def test(dataloader,model,loss_fn):
size = len(dataloader.dataset)
num_batches = len(dataloader)
test_loss, test_acc =0, 0
#当不训练时,梯度停止更新
with torch.no_grad():
for imgs,target in dataloader:
imgs,target = imgs.to(device),target.to(device)
#计算loss
target_pred = model(imgs)
loss = loss_fn(target_pred,target)
test_loss += loss.item()
test_acc += (target_pred.argmax(1) == target).type(torch.float).sum().item()
test_acc /= size
test_loss /= num_batches
return test_acc, test_loss
3.正式训练&设置超参数&保存最佳模型
import time
import copy
'''设置超参数'''
start_epoch = 0
epochs = 50
learn_rate = 1e-3
loss_fn = nn.CrossEntropyLoss() #创建损失函数
optimizer = torch.optim.Adam(model.parameters(), lr=learn_rate)
type(optimizer)
#调用官方动态学习率接口时使用
lambda1 = lambda epoch: 0.92 ** (epoch // 4)
scheduler = torch.optim.lr_scheduler.LambdaLR(optimizer,lr_lambda=lambda1)
train_loss = []
train_acc = []
test_loss = []
test_acc = []
epoch_best_acc = 0
'''加载之前保存的模型'''
if not os.path.exists(output) or not os.path.isdir(output):
os.makedirs(output)
if start_epoch > 0:
resumeFile = os.path.join(output, 'epoch'+str(start_epoch)+'.pkl')
if not os.path.exists(resumeFile) or not os.path.isfile(resumeFile):
start_epoch = 0
else:
model.load_state_dict(torch.load(resumeFile))
'''开始训练模型'''
print('\nStart training...')
best_model = None
for epoch in range(start_epoch, epochs):
model.train()
epoch_train_acc,epoch_train_loss = train(train_dl,model,optimizer,loss_fn)
scheduler.step()
model.eval()
epoch_test_acc, epoch_test_loss = test(test_dl, model, loss_fn)
train_acc.append(epoch_train_acc)
train_loss.append(epoch_train_loss)
test_acc.append(epoch_test_acc)
test_loss.append(epoch_test_loss)
lr = optimizer.state_dict()['param_groups'][0]['lr']
template = ('Epoch:{:2d}, Train_acc:{:.1f}%, Train_loss:{:.3f}, Test_acc:{:.1f}%, Test_loss:{:.3f}, Lr:{:.2E}')
print(time.strftime('[%Y-%m-%d %H:%M:%S]'), template.format(epoch+1, epoch_train_acc*100, epoch_train_loss, epoch_test_acc*100, epoch_test_loss, lr))
# 保存最佳模型
if epoch_test_acc>epoch_best_acc:
''' 保存最优模型参数 '''
epoch_best_acc = epoch_test_acc
best_model = copy.deepcopy(model)
print(('acc = {:.1f}%, saving model to best.pkl').format(epoch_best_acc*100))
saveFile = os.path.join(output, 'best.pkl')
torch.save(best_model.state_dict(), saveFile)
print('Done\n')
''' 保存最新模型参数 '''
saveFile = os.path.join(output, 'epoch'+str(epochs)+'.pkl')
torch.save(model.state_dict(), saveFile)
第一次做的时候,报了很多次错
Start training...
---------------------------------------------------------------------------
TypeError Traceback (most recent call last)
Cell In[7], line 35
32 for epoch in range(start_epoch, epochs):
34 model.train()
---> 35 epoch_train_acc,epoch_train_loss = train(train_dl, model, loss_fn, optimizer)
36 scheduler.step()
38 model.eval()Cell In[5], line 13, in train(dataloader, model, optimizer, loss_fn)
11 #计算误差
12 pred = model(X)
---> 13 loss = loss_fn(pred, y)
15 #反向传播
16 optimizer.zero_grad()TypeError: 'Adam' object is not callable
但是实际我并没有将优化器传给train函数,不过我改了一个模型的参数位置就不报这个错了? 这个error我还没理解,就自己解决了。。。
第二个报错:
解决办法:我把batch_size改成了16,但是依然不行,最后无奈改到4,但是训练了500轮都没办法收敛,所以我利用了一些资源,换到了3090的显卡上跑,实力不够。。。硬件来凑。。。
以下是我用Adam优化器的结果
Start training... [2023-11-30 18:08:23] Epoch: 1, Train_acc:93.1%, Train_loss:0.334, Test_acc:96.7%, Test_loss:0.094, Lr:1.00E-03 acc = 96.7%, saving model to best.pkl [2023-11-30 18:08:34] Epoch: 2, Train_acc:98.2%, Train_loss:0.049, Test_acc:97.5%, Test_loss:0.098, Lr:1.00E-03 acc = 97.5%, saving model to best.pkl [2023-11-30 18:08:45] Epoch: 3, Train_acc:98.5%, Train_loss:0.044, Test_acc:96.2%, Test_loss:0.165, Lr:1.00E-03 [2023-11-30 18:08:55] Epoch: 4, Train_acc:97.6%, Train_loss:0.092, Test_acc:97.5%, Test_loss:0.085, Lr:9.20E-04 [2023-11-30 18:09:05] Epoch: 5, Train_acc:98.2%, Train_loss:0.069, Test_acc:95.8%, Test_loss:0.076, Lr:9.20E-04 [2023-11-30 18:09:15] Epoch: 6, Train_acc:98.9%, Train_loss:0.034, Test_acc:96.7%, Test_loss:0.070, Lr:9.20E-04 [2023-11-30 18:09:25] Epoch: 7, Train_acc:97.5%, Train_loss:0.058, Test_acc:97.9%, Test_loss:0.058, Lr:9.20E-04 acc = 97.9%, saving model to best.pkl [2023-11-30 18:09:39] Epoch: 8, Train_acc:98.1%, Train_loss:0.074, Test_acc:94.2%, Test_loss:0.216, Lr:8.46E-04 [2023-11-30 18:09:51] Epoch: 9, Train_acc:98.1%, Train_loss:0.065, Test_acc:96.7%, Test_loss:0.138, Lr:8.46E-04 [2023-11-30 18:10:01] Epoch:10, Train_acc:97.7%, Train_loss:0.109, Test_acc:95.4%, Test_loss:0.127, Lr:8.46E-04 [2023-11-30 18:10:11] Epoch:11, Train_acc:97.0%, Train_loss:0.165, Test_acc:92.1%, Test_loss:2.717, Lr:8.46E-04 [2023-11-30 18:10:21] Epoch:12, Train_acc:90.8%, Train_loss:0.402, Test_acc:97.5%, Test_loss:0.108, Lr:7.79E-04 [2023-11-30 18:10:31] Epoch:13, Train_acc:95.1%, Train_loss:0.139, Test_acc:95.0%, Test_loss:0.173, Lr:7.79E-04 [2023-11-30 18:10:41] Epoch:14, Train_acc:96.0%, Train_loss:0.111, Test_acc:97.5%, Test_loss:0.075, Lr:7.79E-04 [2023-11-30 18:10:51] Epoch:15, Train_acc:93.2%, Train_loss:0.369, Test_acc:86.2%, Test_loss:0.349, Lr:7.79E-04 [2023-11-30 18:11:01] Epoch:16, Train_acc:94.6%, Train_loss:0.160, Test_acc:95.8%, Test_loss:0.132, Lr:7.16E-04 [2023-11-30 18:11:11] Epoch:17, Train_acc:96.2%, Train_loss:0.117, Test_acc:94.6%, Test_loss:0.168, Lr:7.16E-04 [2023-11-30 18:11:21] Epoch:18, Train_acc:95.8%, Train_loss:0.121, Test_acc:97.9%, Test_loss:0.064, Lr:7.16E-04 [2023-11-30 18:11:31] Epoch:19, Train_acc:97.7%, Train_loss:0.074, Test_acc:97.9%, Test_loss:0.049, Lr:7.16E-04 [2023-11-30 18:11:41] Epoch:20, Train_acc:97.6%, Train_loss:0.062, Test_acc:96.2%, Test_loss:0.090, Lr:6.59E-04 [2023-11-30 18:11:51] Epoch:21, Train_acc:99.3%, Train_loss:0.028, Test_acc:98.3%, Test_loss:0.051, Lr:6.59E-04 acc = 98.3%, saving model to best.pkl [2023-11-30 18:12:02] Epoch:22, Train_acc:99.0%, Train_loss:0.022, Test_acc:97.9%, Test_loss:0.055, Lr:6.59E-04 [2023-11-30 18:12:12] Epoch:23, Train_acc:99.1%, Train_loss:0.024, Test_acc:97.5%, Test_loss:0.080, Lr:6.59E-04 [2023-11-30 18:12:22] Epoch:24, Train_acc:99.8%, Train_loss:0.006, Test_acc:97.9%, Test_loss:0.076, Lr:6.06E-04 [2023-11-30 18:12:32] Epoch:25, Train_acc:99.0%, Train_loss:0.038, Test_acc:92.1%, Test_loss:0.394, Lr:6.06E-04 [2023-11-30 18:12:42] Epoch:26, Train_acc:94.9%, Train_loss:0.159, Test_acc:97.5%, Test_loss:0.081, Lr:6.06E-04 [2023-11-30 18:12:52] Epoch:27, Train_acc:98.0%, Train_loss:0.054, Test_acc:96.2%, Test_loss:0.117, Lr:6.06E-04 [2023-11-30 18:13:02] Epoch:28, Train_acc:99.3%, Train_loss:0.022, Test_acc:95.8%, Test_loss:0.071, Lr:5.58E-04 [2023-11-30 18:13:13] Epoch:29, Train_acc:99.1%, Train_loss:0.020, Test_acc:98.8%, Test_loss:0.073, Lr:5.58E-04 acc = 98.8%, saving model to best.pkl [2023-11-30 18:13:23] Epoch:30, Train_acc:99.4%, Train_loss:0.018, Test_acc:97.1%, Test_loss:0.171, Lr:5.58E-04 [2023-11-30 18:13:33] Epoch:31, Train_acc:99.8%, Train_loss:0.012, Test_acc:99.2%, Test_loss:0.039, Lr:5.58E-04 acc = 99.2%, saving model to best.pkl [2023-11-30 18:13:45] Epoch:32, Train_acc:99.9%, Train_loss:0.004, Test_acc:98.8%, Test_loss:0.044, Lr:5.13E-04 [2023-11-30 18:13:55] Epoch:33, Train_acc:99.9%, Train_loss:0.002, Test_acc:98.3%, Test_loss:0.041, Lr:5.13E-04 [2023-11-30 18:14:05] Epoch:34, Train_acc:100.0%, Train_loss:0.001, Test_acc:98.3%, Test_loss:0.052, Lr:5.13E-04 [2023-11-30 18:14:15] Epoch:35, Train_acc:100.0%, Train_loss:0.000, Test_acc:98.8%, Test_loss:0.046, Lr:5.13E-04 [2023-11-30 18:14:26] Epoch:36, Train_acc:100.0%, Train_loss:0.000, Test_acc:98.3%, Test_loss:0.051, Lr:4.72E-04 [2023-11-30 18:14:36] Epoch:37, Train_acc:100.0%, Train_loss:0.000, Test_acc:98.3%, Test_loss:0.052, Lr:4.72E-04 [2023-11-30 18:14:46] Epoch:38, Train_acc:100.0%, Train_loss:0.000, Test_acc:98.3%, Test_loss:0.067, Lr:4.72E-04 [2023-11-30 18:14:57] Epoch:39, Train_acc:100.0%, Train_loss:0.000, Test_acc:98.3%, Test_loss:0.059, Lr:4.72E-04 [2023-11-30 18:15:07] Epoch:40, Train_acc:100.0%, Train_loss:0.000, Test_acc:98.3%, Test_loss:0.061, Lr:4.34E-04 [2023-11-30 18:15:17] Epoch:41, Train_acc:100.0%, Train_loss:0.000, Test_acc:98.8%, Test_loss:0.078, Lr:4.34E-04 [2023-11-30 18:15:27] Epoch:42, Train_acc:100.0%, Train_loss:0.000, Test_acc:98.3%, Test_loss:0.072, Lr:4.34E-04 [2023-11-30 18:15:37] Epoch:43, Train_acc:100.0%, Train_loss:0.000, Test_acc:98.3%, Test_loss:0.068, Lr:4.34E-04 [2023-11-30 18:15:48] Epoch:44, Train_acc:100.0%, Train_loss:0.000, Test_acc:98.3%, Test_loss:0.070, Lr:4.00E-04 [2023-11-30 18:15:58] Epoch:45, Train_acc:100.0%, Train_loss:0.000, Test_acc:98.3%, Test_loss:0.072, Lr:4.00E-04 [2023-11-30 18:16:08] Epoch:46, Train_acc:100.0%, Train_loss:0.000, Test_acc:98.3%, Test_loss:0.074, Lr:4.00E-04 [2023-11-30 18:16:18] Epoch:47, Train_acc:100.0%, Train_loss:0.000, Test_acc:98.3%, Test_loss:0.083, Lr:4.00E-04 [2023-11-30 18:16:28] Epoch:48, Train_acc:100.0%, Train_loss:0.000, Test_acc:98.3%, Test_loss:0.077, Lr:3.68E-04 [2023-11-30 18:16:39] Epoch:49, Train_acc:100.0%, Train_loss:0.000, Test_acc:98.3%, Test_loss:0.078, Lr:3.68E-04 [2023-11-30 18:16:49] Epoch:50, Train_acc:100.0%, Train_loss:0.000, Test_acc:98.3%, Test_loss:0.080, Lr:3.68E-04 Done
我再换了SGD优化器试了试
四、结果可视化
''' 结果可视化 '''
def displayResult(train_acc, test_acc, train_loss, test_loss, start_epoch, epochs, output=''):
plt.rcParams['font.sans-serif'] = ['SimHei'] # 用来正常显示中文标签
plt.rcParams['axes.unicode_minus'] = False # 用来正常显示负号
plt.rcParams['figure.dpi'] = 100 # 分辨率
epochs_range = range(start_epoch, epochs)
plt.figure('Result Visualization', figsize=(12, 3))
plt.subplot(1, 2, 1)
plt.plot(epochs_range, train_acc, label='Training Accuracy')
plt.plot(epochs_range, test_acc, label='Test Accuracy')
plt.legend(loc='lower right')
plt.title('Training and Validation Accuracy')
plt.subplot(1, 2, 2)
plt.plot(epochs_range, train_loss, label='Training Loss')
plt.plot(epochs_range, test_loss, label='Test Loss')
plt.legend(loc='upper right')
plt.title('Training and Validation Loss')
plt.savefig(os.path.join(output, 'AccuracyLoss.png'))
plt.show()
''' 绘制准确率&损失率曲线图 '''
displayResult(train_acc, test_acc, train_loss, test_loss, start_epoch, epochs, output)
五、加载模型并预测指定图片
''' 预测函数 '''
import PIL
from PIL import Image
def predict(model, img_path):
img = Image.open(img_path)
test_transforms = torchvision.transforms.Compose([
torchvision.transforms.Resize([224, 224]), # 将输入图片resize成统一尺寸
torchvision.transforms.ToTensor(), # 将PIL Image或numpy.ndarray转换为tensor,并归一化到[0,1]之间
torchvision.transforms.Normalize( # 标准化处理-->转换为标准正太分布(高斯分布),使模型更容易收敛
mean=[0.485, 0.456, 0.406],
std=[0.229, 0.224, 0.225]) # 其中 mean=[0.485,0.456,0.406]与std=[0.229,0.224,0.225] 从数据集中随机抽样计算得到的。
])
img = test_transforms(img)
img = img.to(device).unsqueeze(0)
output = model(img)
#print(output.argmax(1))
_, indices = torch.max(output, 1)
percentage = torch.nn.functional.softmax(output, dim=1)[0] * 100
perc = percentage[int(indices)].item()
result = classeNames[indices]
print('predicted:', result, perc)
if __name__=='__main__':
classeNames = list({'Dark': 0, 'Green': 1, 'Light': 2, 'Medium': 3})
num_classes = len(classeNames)
device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
print("Using {} device\n".format(device))
model = VGG16().to(device) # 加载自定义的VGG16模型
model.load_state_dict(torch.load(os.path.join(r'C:\Users\ubuntu\output', 'best.pkl')))
model.eval()
img_path = r'G:\49-data\Light\light (1).png'
predict(model, img_path)
Using cuda device predicted: Light 100.0
个人总结
1.去研究了一下VGG-16这个模型,进而有点好奇VGG-19的效果怎么样
2.关于轻量化模型,在block4,block5中讲卷积核换成1*1,减少输入通道数?