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])

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: 0.5946 Acc: 0.7131
val Loss: 0.1963 Acc: 0.9150

Epoch 1/24
----------
train Loss: 0.5236 Acc: 0.7951
val Loss: 0.3190 Acc: 0.8758

Epoch 2/24
----------
train Loss: 0.7331 Acc: 0.7336
val Loss: 0.4101 Acc: 0.8431

Epoch 3/24
----------
train Loss: 0.4788 Acc: 0.8238
val Loss: 0.3282 Acc: 0.8562

Epoch 4/24
----------
train Loss: 0.3801 Acc: 0.8566
val Loss: 0.2587 Acc: 0.8954

Epoch 5/24
----------
train Loss: 0.4893 Acc: 0.7869
val Loss: 0.3095 Acc: 0.8693

Epoch 6/24
----------
train Loss: 0.4565 Acc: 0.8238
val Loss: 0.3063 Acc: 0.9020

Epoch 7/24
----------
train Loss: 0.3094 Acc: 0.8934
val Loss: 0.2922 Acc: 0.9216

Epoch 8/24
----------
train Loss: 0.2848 Acc: 0.8484
val Loss: 0.2768 Acc: 0.9216

Epoch 9/24
----------
train Loss: 0.3263 Acc: 0.8811
val Loss: 0.2736 Acc: 0.9216

Epoch 10/24
----------
train Loss: 0.2746 Acc: 0.8934
val Loss: 0.2726 Acc: 0.9216

Epoch 11/24
----------
train Loss: 0.2827 Acc: 0.8689
val Loss: 0.2629 Acc: 0.9150

Epoch 12/24
----------
train Loss: 0.2983 Acc: 0.8811
val Loss: 0.2694 Acc: 0.9346

Epoch 13/24
----------
train Loss: 0.2703 Acc: 0.8934
val Loss: 0.2550 Acc: 0.9150

Epoch 14/24
----------
train Loss: 0.2051 Acc: 0.9139
val Loss: 0.2518 Acc: 0.9281

Epoch 15/24
----------
train Loss: 0.2312 Acc: 0.9057
val Loss: 0.2475 Acc: 0.9281

Epoch 16/24
----------
train Loss: 0.2613 Acc: 0.8811
val Loss: 0.2486 Acc: 0.9281

Epoch 17/24
----------
train Loss: 0.2413 Acc: 0.9098
val Loss: 0.2507 Acc: 0.9281

Epoch 18/24
----------
train Loss: 0.2209 Acc: 0.8934
val Loss: 0.2565 Acc: 0.9085

Epoch 19/24
----------
train Loss: 0.3105 Acc: 0.8811
val Loss: 0.2526 Acc: 0.9150

Epoch 20/24
----------
train Loss: 0.2923 Acc: 0.8730
val Loss: 0.2738 Acc: 0.9346

Epoch 21/24
----------
train Loss: 0.3281 Acc: 0.8770
val Loss: 0.2402 Acc: 0.9216

Epoch 22/24
----------
train Loss: 0.2968 Acc: 0.8852
val Loss: 0.2398 Acc: 0.9281

Epoch 23/24
----------
train Loss: 0.2347 Acc: 0.8975
val Loss: 0.2424 Acc: 0.9216

Epoch 24/24
----------
train Loss: 0.2852 Acc: 0.8770
val Loss: 0.2560 Acc: 0.9412

Training complete in 0m 53s
Best val Acc: 0.941176

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
# 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: 0.6732 Acc: 0.6230
val Loss: 0.4119 Acc: 0.7974

Epoch 1/24
----------
train Loss: 0.4737 Acc: 0.7910
val Loss: 0.2613 Acc: 0.9020

Epoch 2/24
----------
train Loss: 0.4814 Acc: 0.7705
val Loss: 0.3054 Acc: 0.8889

Epoch 3/24
----------
train Loss: 0.4801 Acc: 0.7828
val Loss: 0.4398 Acc: 0.8170

Epoch 4/24
----------
train Loss: 0.5135 Acc: 0.7705
val Loss: 0.3538 Acc: 0.8627

Epoch 5/24
----------
train Loss: 0.5238 Acc: 0.7664
val Loss: 0.1921 Acc: 0.9346

Epoch 6/24
----------
train Loss: 0.5178 Acc: 0.7295
val Loss: 0.3205 Acc: 0.8693

Epoch 7/24
----------
train Loss: 0.4640 Acc: 0.8279
val Loss: 0.2157 Acc: 0.9150

Epoch 8/24
----------
train Loss: 0.3189 Acc: 0.8525
val Loss: 0.2196 Acc: 0.9216

Epoch 9/24
----------
train Loss: 0.3761 Acc: 0.8443
val Loss: 0.1715 Acc: 0.9412

Epoch 10/24
----------
train Loss: 0.4566 Acc: 0.7828
val Loss: 0.1815 Acc: 0.9477

Epoch 11/24
----------
train Loss: 0.4128 Acc: 0.8074
val Loss: 0.1791 Acc: 0.9346

Epoch 12/24
----------
train Loss: 0.3742 Acc: 0.8443
val Loss: 0.1814 Acc: 0.9412

Epoch 13/24
----------
train Loss: 0.3514 Acc: 0.8443
val Loss: 0.1773 Acc: 0.9346

Epoch 14/24
----------
train Loss: 0.3364 Acc: 0.8402
val Loss: 0.1723 Acc: 0.9477

Epoch 15/24
----------
train Loss: 0.3743 Acc: 0.8238
val Loss: 0.2025 Acc: 0.9346

Epoch 16/24
----------
train Loss: 0.3806 Acc: 0.8197
val Loss: 0.1929 Acc: 0.9412

Epoch 17/24
----------
train Loss: 0.2910 Acc: 0.8648
val Loss: 0.1858 Acc: 0.9281

Epoch 18/24
----------
train Loss: 0.3100 Acc: 0.8689
val Loss: 0.1654 Acc: 0.9412

Epoch 19/24
----------
train Loss: 0.3245 Acc: 0.8525
val Loss: 0.2089 Acc: 0.9150

Epoch 20/24
----------
train Loss: 0.2960 Acc: 0.8525
val Loss: 0.1790 Acc: 0.9412

Epoch 21/24
----------
train Loss: 0.3195 Acc: 0.8566
val Loss: 0.1754 Acc: 0.9542

Epoch 22/24
----------
train Loss: 0.2925 Acc: 0.8648
val Loss: 0.1806 Acc: 0.9477

Epoch 23/24
----------
train Loss: 0.3474 Acc: 0.8689
val Loss: 0.2028 Acc: 0.9216

Epoch 24/24
----------
train Loss: 0.3815 Acc: 0.8238
val Loss: 0.1897 Acc: 0.9412

Training complete in 0m 34s
Best val Acc: 0.954248

visualize_model(model_conv)

plt.ioff()
plt.show()

Total running time of the script: ( 1 minutes 33.371 seconds)