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 this 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 looks 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
from torch.autograd import Variable
import numpy as np
import torchvision
from torchvision import datasets, models, transforms
import matplotlib.pyplot as plt
import time
import os

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.RandomSizedCrop(224),
        transforms.RandomHorizontalFlip(),
        transforms.ToTensor(),
        transforms.Normalize([0.485, 0.456, 0.406], [0.229, 0.224, 0.225])
    ]),
    'val': transforms.Compose([
        transforms.Scale(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

use_gpu = torch.cuda.is_available()

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 = 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(True)  # Set model to training mode
            else:
                model.train(False)  # Set model to evaluate mode

            running_loss = 0.0
            running_corrects = 0

            # Iterate over data.
            for data in dataloaders[phase]:
                # get the inputs
                inputs, labels = data

                # wrap them in Variable
                if use_gpu:
                    inputs = Variable(inputs.cuda())
                    labels = Variable(labels.cuda())
                else:
                    inputs, labels = Variable(inputs), Variable(labels)

                # zero the parameter gradients
                optimizer.zero_grad()

                # forward
                outputs = model(inputs)
                _, preds = torch.max(outputs.data, 1)
                loss = criterion(outputs, labels)

                # backward + optimize only if in training phase
                if phase == 'train':
                    loss.backward()
                    optimizer.step()

                # statistics
                running_loss += loss.data[0]
                running_corrects += torch.sum(preds == labels.data)

            epoch_loss = running_loss / dataset_sizes[phase]
            epoch_acc = running_corrects / 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 = 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):
    images_so_far = 0
    fig = plt.figure()

    for i, data in enumerate(dataloaders['val']):
        inputs, labels = data
        if use_gpu:
            inputs, labels = Variable(inputs.cuda()), Variable(labels.cuda())
        else:
            inputs, labels = Variable(inputs), Variable(labels)

        outputs = model(inputs)
        _, preds = torch.max(outputs.data, 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:
                return

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)

if use_gpu:
    model_ft = model_ft.cuda()

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: 0.1453 Acc: 0.6844
val Loss: 0.0714 Acc: 0.9085

Epoch 1/24
----------
train Loss: 0.1451 Acc: 0.7582
val Loss: 0.1483 Acc: 0.7451

Epoch 2/24
----------
train Loss: 0.1511 Acc: 0.7828
val Loss: 0.1119 Acc: 0.8235

Epoch 3/24
----------
train Loss: 0.1166 Acc: 0.8033
val Loss: 0.1428 Acc: 0.8039

Epoch 4/24
----------
train Loss: 0.1441 Acc: 0.7746
val Loss: 0.0807 Acc: 0.9020

Epoch 5/24
----------
train Loss: 0.1381 Acc: 0.7951
val Loss: 0.1139 Acc: 0.8693

Epoch 6/24
----------
train Loss: 0.1206 Acc: 0.8279
val Loss: 0.1170 Acc: 0.8562

Epoch 7/24
----------
train Loss: 0.1102 Acc: 0.8525
val Loss: 0.0970 Acc: 0.8497

Epoch 8/24
----------
train Loss: 0.0822 Acc: 0.8648
val Loss: 0.0830 Acc: 0.9150

Epoch 9/24
----------
train Loss: 0.0800 Acc: 0.8811
val Loss: 0.0817 Acc: 0.9281

Epoch 10/24
----------
train Loss: 0.1138 Acc: 0.8074
val Loss: 0.0729 Acc: 0.9020

Epoch 11/24
----------
train Loss: 0.0730 Acc: 0.8648
val Loss: 0.0726 Acc: 0.8954

Epoch 12/24
----------
train Loss: 0.0670 Acc: 0.8689
val Loss: 0.0753 Acc: 0.9216

Epoch 13/24
----------
train Loss: 0.0937 Acc: 0.8525
val Loss: 0.0735 Acc: 0.8889

Epoch 14/24
----------
train Loss: 0.0614 Acc: 0.9016
val Loss: 0.0706 Acc: 0.9216

Epoch 15/24
----------
train Loss: 0.0786 Acc: 0.8402
val Loss: 0.0656 Acc: 0.9216

Epoch 16/24
----------
train Loss: 0.0911 Acc: 0.8443
val Loss: 0.0702 Acc: 0.9020

Epoch 17/24
----------
train Loss: 0.0775 Acc: 0.8770
val Loss: 0.0781 Acc: 0.8758

Epoch 18/24
----------
train Loss: 0.0940 Acc: 0.8443
val Loss: 0.0686 Acc: 0.9085

Epoch 19/24
----------
train Loss: 0.0799 Acc: 0.8689
val Loss: 0.0670 Acc: 0.9281

Epoch 20/24
----------
train Loss: 0.0684 Acc: 0.9057
val Loss: 0.0701 Acc: 0.9150

Epoch 21/24
----------
train Loss: 0.0715 Acc: 0.8730
val Loss: 0.0683 Acc: 0.9216

Epoch 22/24
----------
train Loss: 0.0550 Acc: 0.9139
val Loss: 0.0665 Acc: 0.9150

Epoch 23/24
----------
train Loss: 0.0622 Acc: 0.9016
val Loss: 0.0683 Acc: 0.9020

Epoch 24/24
----------
train Loss: 0.0666 Acc: 0.9098
val Loss: 0.0684 Acc: 0.9216

Training complete in 1m 12s
Best val Acc: 0.928105
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)

if use_gpu:
    model_conv = model_conv.cuda()

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: 0.1844 Acc: 0.5943
val Loss: 0.0651 Acc: 0.9085

Epoch 1/24
----------
train Loss: 0.1425 Acc: 0.7377
val Loss: 0.0610 Acc: 0.8954

Epoch 2/24
----------
train Loss: 0.1050 Acc: 0.8320
val Loss: 0.0554 Acc: 0.9346

Epoch 3/24
----------
train Loss: 0.1659 Acc: 0.7295
val Loss: 0.0726 Acc: 0.8954

Epoch 4/24
----------
train Loss: 0.1292 Acc: 0.8033
val Loss: 0.0585 Acc: 0.9150

Epoch 5/24
----------
train Loss: 0.1208 Acc: 0.7828
val Loss: 0.0443 Acc: 0.9281

Epoch 6/24
----------
train Loss: 0.1085 Acc: 0.8484
val Loss: 0.0478 Acc: 0.9346

Epoch 7/24
----------
train Loss: 0.0929 Acc: 0.8402
val Loss: 0.0442 Acc: 0.9477

Epoch 8/24
----------
train Loss: 0.0844 Acc: 0.8443
val Loss: 0.0520 Acc: 0.9216

Epoch 9/24
----------
train Loss: 0.0796 Acc: 0.8852
val Loss: 0.0671 Acc: 0.9477

Epoch 10/24
----------
train Loss: 0.1070 Acc: 0.8074
val Loss: 0.0462 Acc: 0.9346

Epoch 11/24
----------
train Loss: 0.0851 Acc: 0.8484
val Loss: 0.0528 Acc: 0.9346

Epoch 12/24
----------
train Loss: 0.0748 Acc: 0.8607
val Loss: 0.0434 Acc: 0.9477

Epoch 13/24
----------
train Loss: 0.0814 Acc: 0.8730
val Loss: 0.0487 Acc: 0.9281

Epoch 14/24
----------
train Loss: 0.0848 Acc: 0.8484
val Loss: 0.0496 Acc: 0.9477

Epoch 15/24
----------
train Loss: 0.0893 Acc: 0.8361
val Loss: 0.0499 Acc: 0.9477

Epoch 16/24
----------
train Loss: 0.0867 Acc: 0.8525
val Loss: 0.0477 Acc: 0.9477

Epoch 17/24
----------
train Loss: 0.0907 Acc: 0.8402
val Loss: 0.0471 Acc: 0.9346

Epoch 18/24
----------
train Loss: 0.0822 Acc: 0.8402
val Loss: 0.0502 Acc: 0.9346

Epoch 19/24
----------
train Loss: 0.1005 Acc: 0.8074
val Loss: 0.0439 Acc: 0.9477

Epoch 20/24
----------
train Loss: 0.0784 Acc: 0.8443
val Loss: 0.0435 Acc: 0.9477

Epoch 21/24
----------
train Loss: 0.0834 Acc: 0.8689
val Loss: 0.0518 Acc: 0.9412

Epoch 22/24
----------
train Loss: 0.0682 Acc: 0.8648
val Loss: 0.0463 Acc: 0.9477

Epoch 23/24
----------
train Loss: 0.0691 Acc: 0.9016
val Loss: 0.0519 Acc: 0.9346

Epoch 24/24
----------
train Loss: 0.0700 Acc: 0.8648
val Loss: 0.0489 Acc: 0.9477

Training complete in 0m 51s
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 17.023 seconds)

Gallery generated by Sphinx-Gallery