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']}
dataloders = {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
    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(dataloders['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 dataloders[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(dataloders['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: 0.1250 Acc: 0.7377
val Loss: 0.0615 Acc: 0.9020

Epoch 1/24
----------
train Loss: 0.0979 Acc: 0.8484
val Loss: 0.0575 Acc: 0.9216

Epoch 2/24
----------
train Loss: 0.1657 Acc: 0.7459
val Loss: 0.1188 Acc: 0.8431

Epoch 3/24
----------
train Loss: 0.1010 Acc: 0.8279
val Loss: 0.0695 Acc: 0.9020

Epoch 4/24
----------
train Loss: 0.1266 Acc: 0.7992
val Loss: 0.0833 Acc: 0.8824

Epoch 5/24
----------
train Loss: 0.1336 Acc: 0.7869
val Loss: 0.0786 Acc: 0.9150

Epoch 6/24
----------
train Loss: 0.1154 Acc: 0.8607
val Loss: 0.0592 Acc: 0.9281

Epoch 7/24
----------
train Loss: 0.0873 Acc: 0.8443
val Loss: 0.0523 Acc: 0.9346

Epoch 8/24
----------
train Loss: 0.1066 Acc: 0.8361
val Loss: 0.0509 Acc: 0.9412

Epoch 9/24
----------
train Loss: 0.0706 Acc: 0.8975
val Loss: 0.0473 Acc: 0.9412

Epoch 10/24
----------
train Loss: 0.0955 Acc: 0.8484
val Loss: 0.0464 Acc: 0.9477

Epoch 11/24
----------
train Loss: 0.0807 Acc: 0.8443
val Loss: 0.0583 Acc: 0.9216

Epoch 12/24
----------
train Loss: 0.0699 Acc: 0.8770
val Loss: 0.0526 Acc: 0.9346

Epoch 13/24
----------
train Loss: 0.0764 Acc: 0.9016
val Loss: 0.0472 Acc: 0.9281

Epoch 14/24
----------
train Loss: 0.0604 Acc: 0.8934
val Loss: 0.0483 Acc: 0.9281

Epoch 15/24
----------
train Loss: 0.0583 Acc: 0.9057
val Loss: 0.0504 Acc: 0.9281

Epoch 16/24
----------
train Loss: 0.0530 Acc: 0.9016
val Loss: 0.0508 Acc: 0.9281

Epoch 17/24
----------
train Loss: 0.0898 Acc: 0.8525
val Loss: 0.0467 Acc: 0.9346

Epoch 18/24
----------
train Loss: 0.0540 Acc: 0.9180
val Loss: 0.0454 Acc: 0.9412

Epoch 19/24
----------
train Loss: 0.0635 Acc: 0.8893
val Loss: 0.0449 Acc: 0.9412

Epoch 20/24
----------
train Loss: 0.0529 Acc: 0.9098
val Loss: 0.0483 Acc: 0.9346

Epoch 21/24
----------
train Loss: 0.0554 Acc: 0.8934
val Loss: 0.0447 Acc: 0.9412

Epoch 22/24
----------
train Loss: 0.0645 Acc: 0.9016
val Loss: 0.0442 Acc: 0.9477

Epoch 23/24
----------
train Loss: 0.0679 Acc: 0.8811
val Loss: 0.0417 Acc: 0.9542

Epoch 24/24
----------
train Loss: 0.0662 Acc: 0.8893
val Loss: 0.0487 Acc: 0.9346

Training complete in 0m 46s
Best val Acc: 0.954248
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: 0.1535 Acc: 0.6721
val Loss: 0.0847 Acc: 0.8431

Epoch 1/24
----------
train Loss: 0.1218 Acc: 0.7787
val Loss: 0.1032 Acc: 0.8039

Epoch 2/24
----------
train Loss: 0.1206 Acc: 0.7910
val Loss: 0.0428 Acc: 0.9542

Epoch 3/24
----------
train Loss: 0.1238 Acc: 0.7869
val Loss: 0.0518 Acc: 0.9216

Epoch 4/24
----------
train Loss: 0.1369 Acc: 0.7623
val Loss: 0.0468 Acc: 0.9216

Epoch 5/24
----------
train Loss: 0.1141 Acc: 0.7869
val Loss: 0.0891 Acc: 0.8366

Epoch 6/24
----------
train Loss: 0.0959 Acc: 0.8156
val Loss: 0.0417 Acc: 0.9477

Epoch 7/24
----------
train Loss: 0.0703 Acc: 0.8566
val Loss: 0.0359 Acc: 0.9608

Epoch 8/24
----------
train Loss: 0.0636 Acc: 0.8852
val Loss: 0.0439 Acc: 0.9412

Epoch 9/24
----------
train Loss: 0.0912 Acc: 0.8361
val Loss: 0.0426 Acc: 0.9477

Epoch 10/24
----------
train Loss: 0.0972 Acc: 0.8197
val Loss: 0.0535 Acc: 0.9412

Epoch 11/24
----------
train Loss: 0.0837 Acc: 0.8648
val Loss: 0.0400 Acc: 0.9542

Epoch 12/24
----------
train Loss: 0.0923 Acc: 0.8402
val Loss: 0.0505 Acc: 0.9346

Epoch 13/24
----------
train Loss: 0.0879 Acc: 0.8443
val Loss: 0.0408 Acc: 0.9542

Epoch 14/24
----------
train Loss: 0.0749 Acc: 0.8566
val Loss: 0.0439 Acc: 0.9477

Epoch 15/24
----------
train Loss: 0.0807 Acc: 0.8361
val Loss: 0.0440 Acc: 0.9412

Epoch 16/24
----------
train Loss: 0.0838 Acc: 0.8566
val Loss: 0.0464 Acc: 0.9477

Epoch 17/24
----------
train Loss: 0.0853 Acc: 0.8443
val Loss: 0.0410 Acc: 0.9608

Epoch 18/24
----------
train Loss: 0.0750 Acc: 0.8852
val Loss: 0.0428 Acc: 0.9542

Epoch 19/24
----------
train Loss: 0.0905 Acc: 0.8566
val Loss: 0.0507 Acc: 0.9281

Epoch 20/24
----------
train Loss: 0.0871 Acc: 0.8525
val Loss: 0.0395 Acc: 0.9542

Epoch 21/24
----------
train Loss: 0.0866 Acc: 0.8402
val Loss: 0.0437 Acc: 0.9477

Epoch 22/24
----------
train Loss: 0.0805 Acc: 0.8279
val Loss: 0.0503 Acc: 0.9412

Epoch 23/24
----------
train Loss: 0.0792 Acc: 0.8484
val Loss: 0.0446 Acc: 0.9412

Epoch 24/24
----------
train Loss: 0.0736 Acc: 0.8443
val Loss: 0.0424 Acc: 0.9412

Training complete in 0m 36s
Best val Acc: 0.960784
visualize_model(model_conv)

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

Total running time of the script: ( 1 minutes 30.495 seconds)

Generated by Sphinx-Gallery