Shortcuts

Transfer Learning Tutorial

Author: Sasank Chilamkurthy

In this tutorial, you will learn how to train your network using transfer learning. You can read more about the transfer learning at cs231n notes

Quoting these notes,

In practice, very few people train an entire Convolutional Network from scratch (with random initialization), because it is relatively rare to have a dataset of sufficient size. Instead, it is common to pretrain a ConvNet on a very large dataset (e.g. ImageNet, which contains 1.2 million images with 1000 categories), and then use the ConvNet either as an initialization or a fixed feature extractor for the task of interest.

These two major transfer learning scenarios look as follows:

  • Finetuning the convnet: Instead of random initializaion, we initialize the network with a pretrained network, like the one that is trained on imagenet 1000 dataset. Rest of the training looks as usual.
  • ConvNet as fixed feature extractor: Here, we will freeze the weights for all of the network except that of the final fully connected layer. This last fully connected layer is replaced with a new one with random weights and only this layer is trained.
# 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 numpy as np
import torchvision
from torchvision import datasets, models, transforms
import matplotlib.pyplot as plt
import time
import os
import copy

plt.ion()   # interactive mode

Load Data

We will use torchvision and torch.utils.data packages for loading the data.

The problem we’re going to solve today is to train a model to classify ants and bees. We have about 120 training images each for ants and bees. There are 75 validation images for each class. Usually, this is a very small dataset to generalize upon, if trained from scratch. Since we are using transfer learning, we should be able to generalize reasonably well.

This dataset is a very small subset of imagenet.

Note

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 = '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")

Visualize a few images

Let’s visualize a few training images so as to understand the data augmentations.

def imshow(inp, title=None):
    """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
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])
../_images/sphx_glr_transfer_learning_tutorial_001.png

Training the model

Now, let’s write a general function to train a model. Here, we will illustrate:

  • Scheduling the learning rate
  • Saving the best model

In the following, parameter scheduler is an LR scheduler object from torch.optim.lr_scheduler.

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('Epoch {}/{}'.format(epoch, num_epochs - 1))
        print('-' * 10)

        # Each epoch has a training and validation phase
        for phase in ['train', 'val']:
            if phase == 'train':
                scheduler.step()
                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)

            epoch_loss = running_loss / dataset_sizes[phase]
            epoch_acc = running_corrects.double() / dataset_sizes[phase]

            print('{} Loss: {:.4f} Acc: {:.4f}'.format(
                phase, epoch_loss, epoch_acc))

            # 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('Training complete in {:.0f}m {:.0f}s'.format(
        time_elapsed // 60, time_elapsed % 60))
    print('Best val Acc: {:4f}'.format(best_acc))

    # load best model weights
    model.load_state_dict(best_model_wts)
    return model

Visualizing the model predictions

Generic function to display predictions for a few images

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('predicted: {}'.format(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)

Finetuning the convnet

Load a pretrained model and reset final fully connected layer.

model_ft = models.resnet18(pretrained=True)
num_ftrs = model_ft.fc.in_features
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)

Train and evaluate

It should take around 15-25 min on CPU. On GPU though, it takes less than a minute.

model_ft = train_model(model_ft, criterion, optimizer_ft, exp_lr_scheduler,
                       num_epochs=25)

Out:

Epoch 0/24
----------
train Loss: 0.6370 Acc: 0.6762
val Loss: 0.2259 Acc: 0.9281

Epoch 1/24
----------
train Loss: 0.5249 Acc: 0.7787
val Loss: 0.3082 Acc: 0.8824

Epoch 2/24
----------
train Loss: 0.4280 Acc: 0.8156
val Loss: 0.7138 Acc: 0.7386

Epoch 3/24
----------
train Loss: 0.5680 Acc: 0.7582
val Loss: 0.3947 Acc: 0.8301

Epoch 4/24
----------
train Loss: 0.4301 Acc: 0.8320
val Loss: 0.2369 Acc: 0.8954

Epoch 5/24
----------
train Loss: 0.4234 Acc: 0.8279
val Loss: 0.2678 Acc: 0.8954

Epoch 6/24
----------
train Loss: 0.7623 Acc: 0.7213
val Loss: 1.0202 Acc: 0.6993

Epoch 7/24
----------
train Loss: 0.5394 Acc: 0.8320
val Loss: 0.2270 Acc: 0.9346

Epoch 8/24
----------
train Loss: 0.3026 Acc: 0.8607
val Loss: 0.1838 Acc: 0.9346

Epoch 9/24
----------
train Loss: 0.2919 Acc: 0.8893
val Loss: 0.2263 Acc: 0.9216

Epoch 10/24
----------
train Loss: 0.3342 Acc: 0.8689
val Loss: 0.2096 Acc: 0.9150

Epoch 11/24
----------
train Loss: 0.3549 Acc: 0.8484
val Loss: 0.2193 Acc: 0.9150

Epoch 12/24
----------
train Loss: 0.3048 Acc: 0.8730
val Loss: 0.2074 Acc: 0.9150

Epoch 13/24
----------
train Loss: 0.2064 Acc: 0.9221
val Loss: 0.1688 Acc: 0.9346

Epoch 14/24
----------
train Loss: 0.3431 Acc: 0.8648
val Loss: 0.1909 Acc: 0.9216

Epoch 15/24
----------
train Loss: 0.3488 Acc: 0.8648
val Loss: 0.1709 Acc: 0.9281

Epoch 16/24
----------
train Loss: 0.1918 Acc: 0.9221
val Loss: 0.1769 Acc: 0.9346

Epoch 17/24
----------
train Loss: 0.2416 Acc: 0.8852
val Loss: 0.1971 Acc: 0.9150

Epoch 18/24
----------
train Loss: 0.2959 Acc: 0.8852
val Loss: 0.1812 Acc: 0.9281

Epoch 19/24
----------
train Loss: 0.3297 Acc: 0.8730
val Loss: 0.1725 Acc: 0.9216

Epoch 20/24
----------
train Loss: 0.2402 Acc: 0.8893
val Loss: 0.1698 Acc: 0.9346

Epoch 21/24
----------
train Loss: 0.2934 Acc: 0.8852
val Loss: 0.1717 Acc: 0.9346

Epoch 22/24
----------
train Loss: 0.2310 Acc: 0.9098
val Loss: 0.1866 Acc: 0.9216

Epoch 23/24
----------
train Loss: 0.2914 Acc: 0.8648
val Loss: 0.2050 Acc: 0.9281

Epoch 24/24
----------
train Loss: 0.2861 Acc: 0.8852
val Loss: 0.1816 Acc: 0.9216

Training complete in 1m 14s
Best val Acc: 0.934641
visualize_model(model_ft)
../_images/sphx_glr_transfer_learning_tutorial_002.png

ConvNet as fixed feature extractor

Here, we need to freeze all the network except the final layer. We need to set requires_grad == False to freeze the parameters so that the gradients are not computed in backward().

You can read more about this in the documentation here.

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
# opoosed 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)

Train and evaluate

On CPU this will take about half the time compared to previous scenario. This is expected as gradients don’t need to be computed for most of the network. However, forward does need to be computed.

model_conv = train_model(model_conv, criterion, optimizer_conv,
                         exp_lr_scheduler, num_epochs=25)

Out:

Epoch 0/24
----------
train Loss: 0.5693 Acc: 0.6803
val Loss: 0.2276 Acc: 0.9150

Epoch 1/24
----------
train Loss: 0.4701 Acc: 0.8033
val Loss: 0.1747 Acc: 0.9477

Epoch 2/24
----------
train Loss: 0.4607 Acc: 0.7869
val Loss: 0.1671 Acc: 0.9412

Epoch 3/24
----------
train Loss: 0.4480 Acc: 0.8238
val Loss: 0.2159 Acc: 0.9281

Epoch 4/24
----------
train Loss: 0.4380 Acc: 0.8238
val Loss: 0.1914 Acc: 0.9346

Epoch 5/24
----------
train Loss: 0.4213 Acc: 0.7746
val Loss: 0.2610 Acc: 0.8954

Epoch 6/24
----------
train Loss: 0.5766 Acc: 0.7336
val Loss: 0.2643 Acc: 0.8954

Epoch 7/24
----------
train Loss: 0.3505 Acc: 0.8525
val Loss: 0.1956 Acc: 0.9281

Epoch 8/24
----------
train Loss: 0.4029 Acc: 0.8197
val Loss: 0.1807 Acc: 0.9477

Epoch 9/24
----------
train Loss: 0.3720 Acc: 0.8443
val Loss: 0.2048 Acc: 0.9281

Epoch 10/24
----------
train Loss: 0.3451 Acc: 0.8361
val Loss: 0.1950 Acc: 0.9346

Epoch 11/24
----------
train Loss: 0.2778 Acc: 0.8689
val Loss: 0.2092 Acc: 0.9150

Epoch 12/24
----------
train Loss: 0.3632 Acc: 0.8402
val Loss: 0.2514 Acc: 0.8954

Epoch 13/24
----------
train Loss: 0.3685 Acc: 0.8484
val Loss: 0.1796 Acc: 0.9412

Epoch 14/24
----------
train Loss: 0.3438 Acc: 0.8402
val Loss: 0.1807 Acc: 0.9477

Epoch 15/24
----------
train Loss: 0.3027 Acc: 0.8811
val Loss: 0.2003 Acc: 0.9216

Epoch 16/24
----------
train Loss: 0.3236 Acc: 0.8566
val Loss: 0.1764 Acc: 0.9412

Epoch 17/24
----------
train Loss: 0.3754 Acc: 0.8156
val Loss: 0.1740 Acc: 0.9412

Epoch 18/24
----------
train Loss: 0.4050 Acc: 0.8238
val Loss: 0.1863 Acc: 0.9346

Epoch 19/24
----------
train Loss: 0.3322 Acc: 0.8525
val Loss: 0.2191 Acc: 0.9281

Epoch 20/24
----------
train Loss: 0.2870 Acc: 0.8893
val Loss: 0.1741 Acc: 0.9477

Epoch 21/24
----------
train Loss: 0.3395 Acc: 0.8566
val Loss: 0.1893 Acc: 0.9281

Epoch 22/24
----------
train Loss: 0.3068 Acc: 0.8648
val Loss: 0.1923 Acc: 0.9216

Epoch 23/24
----------
train Loss: 0.3427 Acc: 0.8402
val Loss: 0.1868 Acc: 0.9346

Epoch 24/24
----------
train Loss: 0.2819 Acc: 0.8811
val Loss: 0.1999 Acc: 0.9216

Training complete in 0m 37s
Best val Acc: 0.947712
visualize_model(model_conv)

plt.ioff()
plt.show()
../_images/sphx_glr_transfer_learning_tutorial_003.png

Total running time of the script: ( 2 minutes 0.557 seconds)

Gallery generated by Sphinx-Gallery

Docs

Access comprehensive developer documentation for PyTorch

View Docs

Tutorials

Get in-depth tutorials for beginners and advanced developers

View Tutorials

Resources

Find development resources and get your questions answered

View Resources