https://pytorch.org/tutorials/beginner/transfer_learning_tutorial.html
Learn how to train a CNN network for image classification using transfer learning in this tutorial.
In fact, very few people train an entire convolutional network from scratch (random initialization) because it is relatively rare to have a sufficiently large dataset. Instead, it is common to pre-train a ConvNet on a very large dataset (such as ImageNet, which contains 1.2 million images in 1000 categories), and then use the ConvNet as an initialization or a fixed feature extractor for its own task.
In general there are two main transfer learning scenarios:
Fine-tuning convnet: that is, finetune convnet, use the pre-trained network to initialize the network instead of random initialization , just like using the network trained on the imagenet 1000 dataset. The rest of the training looks as usual.
ConvNet as a fixed feature extractor: freeze all network weights except the final fully connected layer . The last fully connected layer is replaced with a new layer with random weights, and only this layer is trained. [Only the parameters of this layer will be updated during backpropagation]
1. First import the relevant packages
# License: BSD
# Author: Sasank Chilamkurthy
from __future__ import print_function, division
import torch
import torch.nn as nn
import torch.optim as optim
from torch.optim import lr_scheduler
import torch.backends.cudnn as cudnn
import numpy as np
import torchvision
from torchvision import datasets, models, transforms
import matplotlib.pyplot as plt
import time
import os
import copy
cudnn.benchmark = True
plt.ion() # interactive mode2. Load data
Using torchvision and the torch.utils.data package to load the data, the problem to be solved today is to train a model to classify ants and bees. There are about 120 training images for ants and bees. There are 75 validation images for each category. Typically, this is a very small dataset for generalization (i.e. hard to generalize to) if training from scratch. Since we are using transfer learning, it should generalize well. This dataset is a very small subset of imagenet.
Download the data from here and extract it to the current directory.
# Data augmentation and normalization for training
# Just normalization for validation
data_transforms = {
'train': transforms.Compose([
transforms.RandomResizedCrop(224),
transforms.RandomHorizontalFlip(),
transforms.ToTensor(),
transforms.Normalize([0.485, 0.456, 0.406], [0.229, 0.224, 0.225])
]),
'val': transforms.Compose([
transforms.Resize(256),
transforms.CenterCrop(224),
transforms.ToTensor(),
transforms.Normalize([0.485, 0.456, 0.406], [0.229, 0.224, 0.225])
]),
}
data_dir = 'data/hymenoptera_data'
image_datasets = {x: datasets.ImageFolder(os.path.join(data_dir, x),
data_transforms[x])
for x in ['train', 'val']}
dataloaders = {x: torch.utils.data.DataLoader(image_datasets[x], batch_size=4,
shuffle=True, num_workers=4)
for x in ['train', 'val']}
dataset_sizes = {x: len(image_datasets[x]) for x in ['train', 'val']}
class_names = image_datasets['train'].classes
device = torch.device("cuda:0" if torch.cuda.is_available() else "cpu")
I printed image_datasets, and I can see that image_datasets is a structure, including the number of samples, paths, and transform operations of training data and test data.
{'train': Dataset ImageFolder
Number of datapoints: 244
Root location: data/hymenoptera_data\train
StandardTransform
Transform: Compose(
RandomResizedCrop(size=(224, 224), scale=(0.08, 1.0), ratio=(0.75, 1.3333) , interpolation=bilinear), antialias=None)
RandomHorizontalFlip(p=0.5)
ToTensor()
Normalize(mean=[0.485, 0.456, 0.406], std=[0.229, 0.224, 0.225]) )
, 'val': Dataset ImageFolder
Number of datapoints: 153
Root location: data/hymenoptera_data\val
StandardTransform
Transform: Compose(
Resize(size=256, interpolation=bilinear, max_size=None, antialias=None)
CenterCrop(size=(224, 224))
ToTensor()
Normalize(mean=[0.485, 0.456, 0.406], std=[0.229, 0.224, 0.225])
)}
print dataloaders：
{'train': <torch.utils.data.dataloader.DataLoader object at 0x000001940A9E99D0>, 'val': <torch.utils.data.dataloader.DataLoader object at 0x000001940A9E99A0>}
3. Data visualization
def imshow(inp, title=None): # 输出为inp
"""Imshow for Tensor."""
inp = inp.numpy().transpose((1, 2, 0))
mean = np.array([0.485, 0.456, 0.406])
std = np.array([0.229, 0.224, 0.225])
inp = std * inp + mean
inp = np.clip(inp, 0, 1)
plt.imshow(inp)
if title is not None:
plt.title(title)
plt.pause(0.001) # pause a bit so that plots are updated
# Get a batch of training data 得到一个batch=4的train数据
inputs, classes = next(iter(dataloaders['train']))
# Make a grid from batch
out = torchvision.utils.make_grid(inputs)
imshow(out, title=[class_names[x] for x in classes])
4. Training model
Now, let’s write a general function to train a model. Here, we will illustrate:
Scheduling the learning rate
Saving the best model (save the best model)
In the following, parameter scheduler is an LR scheduler object from torch.optim.lr_scheduler
That is, a function that describes the training process.
def train_model(model, criterion, optimizer, scheduler, num_epochs=25):
since = time.time()
best_model_wts = copy.deepcopy(model.state_dict())
best_acc = 0.0
for epoch in range(num_epochs):
print(f'Epoch {epoch}/{num_epochs - 1}')
print('-' * 10)
# Each epoch has a training and validation phase
for phase in ['train', 'val']:
if phase == 'train':
model.train() # Set model to training mode
else:
model.eval() # Set model to evaluate mode
running_loss = 0.0
running_corrects = 0
# Iterate over data.
for inputs, labels in dataloaders[phase]:
inputs = inputs.to(device)
labels = labels.to(device)
# zero the parameter gradients
optimizer.zero_grad()
# forward
# track history if only in train
with torch.set_grad_enabled(phase == 'train'):
outputs = model(inputs)
_, preds = torch.max(outputs, 1)
loss = criterion(outputs, labels)
# backward + optimize only if in training phase
if phase == 'train':
loss.backward()
optimizer.step()
# statistics
running_loss += loss.item() * inputs.size(0)
running_corrects += torch.sum(preds == labels.data)
if phase == 'train':
scheduler.step()
epoch_loss = running_loss / dataset_sizes[phase]
epoch_acc = running_corrects.double() / dataset_sizes[phase]
print(f'{phase} Loss: {epoch_loss:.4f} Acc: {epoch_acc:.4f}')
# deep copy the model
if phase == 'val' and epoch_acc > best_acc:
best_acc = epoch_acc
best_model_wts = copy.deepcopy(model.state_dict())
print()
time_elapsed = time.time() - since
print(f'Training complete in {time_elapsed // 60:.0f}m {time_elapsed % 60:.0f}s')
print(f'Best val Acc: {best_acc:4f}')
# load best model weights
model.load_state_dict(best_model_wts)
return model5. Visualize model prediction results
def visualize_model(model, num_images=6):
was_training = model.training
model.eval()
images_so_far = 0
fig = plt.figure()
with torch.no_grad():
for i, (inputs, labels) in enumerate(dataloaders['val']):
inputs = inputs.to(device)
labels = labels.to(device)
outputs = model(inputs)
_, preds = torch.max(outputs, 1)
for j in range(inputs.size()[0]):
images_so_far += 1
ax = plt.subplot(num_images//2, 2, images_so_far)
ax.axis('off')
ax.set_title(f'predicted: {class_names[preds[j]]}')
imshow(inputs.cpu().data[j])
if images_so_far == num_images:
model.train(mode=was_training)
return
model.train(mode=was_training)6. Finetuning pre-training network
Load a pretrained model and reset final fully connected layer. Fine-tune ConvNet , load the pre-trained model and reset the final fully connected layer, and perform training, mainly focusing on the fifth line model_ft.fc = nn.Linear(num_ftrs, 2) is to classify multiple The network is changed to a binary classification network:
model_ft = models.resnet18(pretrained=True)
num_ftrs = model_ft.fc.in_features
# Here the size of each output sample is set to 2.
# Alternatively, it can be generalized to nn.Linear(num_ftrs, len(class_names)).
model_ft.fc = nn.Linear(num_ftrs, 2)
model_ft = model_ft.to(device)
criterion = nn.CrossEntropyLoss()
# Observe that all parameters are being optimized
optimizer_ft = optim.SGD(model_ft.parameters(), lr=0.001, momentum=0.9)
# Decay LR by a factor of 0.1 every 7 epochs
exp_lr_scheduler = lr_scheduler.StepLR(optimizer_ft, step_size=7, gamma=0.1)七、Train and evaluate
model_ft = train_model(model_ft, criterion, optimizer_ft, exp_lr_scheduler,
num_epochs=25)Epoch 22/24
----------
train Loss: 0.3835 Acc: 0.8320
val Loss: 0.2321 Acc: 0.8954
Epoch 23/24
----------
train Loss: 0.3731 Acc: 0.8320
val Loss: 0.2228 Acc: 0.8954
Epoch 24/24
----------
train Loss: 0.3357 Acc: 0.8689
val Loss: 0.2345 Acc: 0.9020
Training complete in 1m 26s
Best val Acc: 0.928105
visualize_model(model_ft)
ConvNet as fixed feature extractor
Then use ConvNet as a fixed feature extractor, where you need to freeze all networks except the last layer. Freeze parameters by setting requires_grad=Falsebackward() so that their gradients are not computed when backpropagating backward().
model_conv = torchvision.models.resnet18(pretrained=True)
for param in model_conv.parameters():
param.requires_grad = False
# Parameters of newly constructed modules have requires_grad=True by default
num_ftrs = model_conv.fc.in_features
model_conv.fc = nn.Linear(num_ftrs, 2)
model_conv = model_conv.to(device)
criterion = nn.CrossEntropyLoss()
# Observe that only parameters of final layer are being optimized as
# opposed to before.
optimizer_conv = optim.SGD(model_conv.fc.parameters(), lr=0.001, momentum=0.9)
# Decay LR by a factor of 0.1 every 7 epochs
exp_lr_scheduler = lr_scheduler.StepLR(optimizer_conv, step_size=7, gamma=0.1)