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 = '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")

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])

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: 0.5900 Acc: 0.7131
val Loss: 0.2508 Acc: 0.9020

Epoch 1/24
----------
train Loss: 0.6034 Acc: 0.7828
val Loss: 0.3181 Acc: 0.8627

Epoch 2/24
----------
train Loss: 0.6150 Acc: 0.7582
val Loss: 0.4903 Acc: 0.8366

Epoch 3/24
----------
train Loss: 0.6650 Acc: 0.7377
val Loss: 0.6294 Acc: 0.7582

Epoch 4/24
----------
train Loss: 0.4935 Acc: 0.7828
val Loss: 0.2644 Acc: 0.8889

Epoch 5/24
----------
train Loss: 0.3841 Acc: 0.8238
val Loss: 0.2408 Acc: 0.9216

Epoch 6/24
----------
train Loss: 0.5352 Acc: 0.8156
val Loss: 0.2250 Acc: 0.9150

Epoch 7/24
----------
train Loss: 0.2252 Acc: 0.9385
val Loss: 0.1917 Acc: 0.9477

Epoch 8/24
----------
train Loss: 0.3395 Acc: 0.8197
val Loss: 0.1738 Acc: 0.9477

Epoch 9/24
----------
train Loss: 0.3363 Acc: 0.8607
val Loss: 0.2522 Acc: 0.9216

Epoch 10/24
----------
train Loss: 0.2878 Acc: 0.8607
val Loss: 0.1787 Acc: 0.9412

Epoch 11/24
----------
train Loss: 0.2831 Acc: 0.8770
val Loss: 0.1805 Acc: 0.9346

Epoch 12/24
----------
train Loss: 0.2290 Acc: 0.9016
val Loss: 0.1898 Acc: 0.9412

Epoch 13/24
----------
train Loss: 0.2494 Acc: 0.9016
val Loss: 0.1729 Acc: 0.9412

Epoch 14/24
----------
train Loss: 0.3435 Acc: 0.8689
val Loss: 0.1736 Acc: 0.9412

Epoch 15/24
----------
train Loss: 0.2274 Acc: 0.9057
val Loss: 0.1692 Acc: 0.9542

Epoch 16/24
----------
train Loss: 0.3154 Acc: 0.8689
val Loss: 0.1742 Acc: 0.9412

Epoch 17/24
----------
train Loss: 0.2749 Acc: 0.8893
val Loss: 0.1826 Acc: 0.9412

Epoch 18/24
----------
train Loss: 0.2673 Acc: 0.8770
val Loss: 0.1731 Acc: 0.9281

Epoch 19/24
----------
train Loss: 0.2865 Acc: 0.8730
val Loss: 0.1867 Acc: 0.9346

Epoch 20/24
----------
train Loss: 0.3061 Acc: 0.8648
val Loss: 0.1966 Acc: 0.9477

Epoch 21/24
----------
train Loss: 0.2638 Acc: 0.9016
val Loss: 0.1973 Acc: 0.9477

Epoch 22/24
----------
train Loss: 0.2602 Acc: 0.8893
val Loss: 0.1769 Acc: 0.9281

Epoch 23/24
----------
train Loss: 0.2817 Acc: 0.9016
val Loss: 0.1756 Acc: 0.9412

Epoch 24/24
----------
train Loss: 0.2959 Acc: 0.8730
val Loss: 0.1790 Acc: 0.9281

Training complete in 1m 8s
Best val Acc: 0.954248

visualize_model(model_ft)

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
# 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)

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: 0.6463 Acc: 0.6803
val Loss: 0.1949 Acc: 0.9477

Epoch 1/24
----------
train Loss: 0.4923 Acc: 0.8033
val Loss: 0.1696 Acc: 0.9477

Epoch 2/24
----------
train Loss: 0.4234 Acc: 0.8115
val Loss: 0.4379 Acc: 0.7712

Epoch 3/24
----------
train Loss: 0.5606 Acc: 0.7582
val Loss: 0.6383 Acc: 0.7451

Epoch 4/24
----------
train Loss: 0.7560 Acc: 0.7295
val Loss: 0.1888 Acc: 0.9412

Epoch 5/24
----------
train Loss: 0.4316 Acc: 0.8197
val Loss: 0.1999 Acc: 0.9477

Epoch 6/24
----------
train Loss: 0.7722 Acc: 0.7131
val Loss: 0.1975 Acc: 0.9477

Epoch 7/24
----------
train Loss: 0.3685 Acc: 0.8607
val Loss: 0.2000 Acc: 0.9477

Epoch 8/24
----------
train Loss: 0.2968 Acc: 0.8811
val Loss: 0.1916 Acc: 0.9477

Epoch 9/24
----------
train Loss: 0.3396 Acc: 0.8525
val Loss: 0.2165 Acc: 0.9542

Epoch 10/24
----------
train Loss: 0.3885 Acc: 0.8320
val Loss: 0.2109 Acc: 0.9542

Epoch 11/24
----------
train Loss: 0.4107 Acc: 0.8156
val Loss: 0.1881 Acc: 0.9477

Epoch 12/24
----------
train Loss: 0.3249 Acc: 0.8730
val Loss: 0.1747 Acc: 0.9542

Epoch 13/24
----------
train Loss: 0.3439 Acc: 0.8525
val Loss: 0.1950 Acc: 0.9477

Epoch 14/24
----------
train Loss: 0.3641 Acc: 0.8443
val Loss: 0.1992 Acc: 0.9412

Epoch 15/24
----------
train Loss: 0.3272 Acc: 0.8443
val Loss: 0.2320 Acc: 0.9412

Epoch 16/24
----------
train Loss: 0.3102 Acc: 0.8730
val Loss: 0.1867 Acc: 0.9477

Epoch 17/24
----------
train Loss: 0.4226 Acc: 0.8238
val Loss: 0.1872 Acc: 0.9542

Epoch 18/24
----------
train Loss: 0.3452 Acc: 0.8443
val Loss: 0.1812 Acc: 0.9542

Epoch 19/24
----------
train Loss: 0.3697 Acc: 0.8525
val Loss: 0.1890 Acc: 0.9477

Epoch 20/24
----------
train Loss: 0.3078 Acc: 0.8607
val Loss: 0.1976 Acc: 0.9608

Epoch 21/24
----------
train Loss: 0.3161 Acc: 0.8770
val Loss: 0.1982 Acc: 0.9412

Epoch 22/24
----------
train Loss: 0.3749 Acc: 0.8320
val Loss: 0.2035 Acc: 0.9477

Epoch 23/24
----------
train Loss: 0.3298 Acc: 0.8525
val Loss: 0.1855 Acc: 0.9477

Epoch 24/24
----------
train Loss: 0.3597 Acc: 0.8402
val Loss: 0.1878 Acc: 0.9542

Training complete in 0m 34s
Best val Acc: 0.960784

visualize_model(model_conv)

plt.ioff()
plt.show()

Total running time of the script: ( 1 minutes 55.614 seconds)