Transfer Learning for Computer Vision Tutorial

Author: Sasank Chilamkurthy

In this tutorial, you will learn how to train a convolutional neural network for image classification 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 initialization, 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 torch.backends.cudnn as cudnn
import numpy as np
import torchvision
from torchvision import datasets, models, transforms
import matplotlib.pyplot as plt
import time
import os
from tempfile import TemporaryDirectory

cudnn.benchmark = True
plt.ion()   # interactive mode

<contextlib.ExitStack object at 0x7f0a630c7760>

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):
    """Display image 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()

    # Create a temporary directory to save training checkpoints
    with TemporaryDirectory() as tempdir:
        best_model_params_path = os.path.join(tempdir, 'best_model_params.pt')

        torch.save(model.state_dict(), best_model_params_path)
        best_acc = 0.0

        for epoch in range(num_epochs):
            print(f'Epoch {epoch}/{num_epochs - 1}')
            print('-' * 10)

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

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

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

                # deep copy the model
                if phase == 'val' and epoch_acc > best_acc:
                    best_acc = epoch_acc
                    torch.save(model.state_dict(), best_model_params_path)

            print()

        time_elapsed = time.time() - since
        print(f'Training complete in {time_elapsed // 60:.0f}m {time_elapsed % 60:.0f}s')
        print(f'Best val Acc: {best_acc:4f}')

        # load best model weights
        model.load_state_dict(torch.load(best_model_params_path))
    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(f'predicted: {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(weights='IMAGENET1K_V1')
num_ftrs = model_ft.fc.in_features
# Here the size of each output sample is set to 2.
# Alternatively, it can be generalized to ``nn.Linear(num_ftrs, len(class_names))``.
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)

Downloading: "https://download.pytorch.org/models/resnet18-f37072fd.pth" to /var/lib/jenkins/.cache/torch/hub/checkpoints/resnet18-f37072fd.pth

  0%|          | 0.00/44.7M [00:00<?, ?B/s]
 23%|##3       | 10.5M/44.7M [00:00<00:00, 110MB/s]
 61%|######    | 27.0M/44.7M [00:00<00:00, 147MB/s]
100%|##########| 44.7M/44.7M [00:00<00:00, 162MB/s]

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)

Epoch 0/24
----------
train Loss: 0.6338 Acc: 0.6762
val Loss: 0.2142 Acc: 0.9281

Epoch 1/24
----------
train Loss: 0.5182 Acc: 0.7746
val Loss: 0.2491 Acc: 0.9150

Epoch 2/24
----------
train Loss: 0.6127 Acc: 0.7910
val Loss: 0.4367 Acc: 0.8627

Epoch 3/24
----------
train Loss: 0.4770 Acc: 0.7869
val Loss: 0.4026 Acc: 0.8758

Epoch 4/24
----------
train Loss: 0.4512 Acc: 0.8279
val Loss: 0.3185 Acc: 0.8693

Epoch 5/24
----------
train Loss: 0.5259 Acc: 0.7541
val Loss: 0.5836 Acc: 0.7582

Epoch 6/24
----------
train Loss: 0.6043 Acc: 0.8279
val Loss: 0.3168 Acc: 0.8758

Epoch 7/24
----------
train Loss: 0.5271 Acc: 0.8279
val Loss: 0.2759 Acc: 0.9216

Epoch 8/24
----------
train Loss: 0.2644 Acc: 0.9016
val Loss: 0.2408 Acc: 0.9281

Epoch 9/24
----------
train Loss: 0.4128 Acc: 0.8648
val Loss: 0.3083 Acc: 0.9085

Epoch 10/24
----------
train Loss: 0.3303 Acc: 0.8811
val Loss: 0.2314 Acc: 0.9346

Epoch 11/24
----------
train Loss: 0.3300 Acc: 0.8648
val Loss: 0.2451 Acc: 0.9150

Epoch 12/24
----------
train Loss: 0.2825 Acc: 0.8811
val Loss: 0.2626 Acc: 0.9216

Epoch 13/24
----------
train Loss: 0.2526 Acc: 0.9057
val Loss: 0.2088 Acc: 0.9346

Epoch 14/24
----------
train Loss: 0.2926 Acc: 0.8730
val Loss: 0.2443 Acc: 0.9281

Epoch 15/24
----------
train Loss: 0.2828 Acc: 0.9016
val Loss: 0.2125 Acc: 0.9346

Epoch 16/24
----------
train Loss: 0.3893 Acc: 0.8689
val Loss: 0.2252 Acc: 0.9281

Epoch 17/24
----------
train Loss: 0.2619 Acc: 0.8934
val Loss: 0.2119 Acc: 0.9346

Epoch 18/24
----------
train Loss: 0.2832 Acc: 0.8934
val Loss: 0.2423 Acc: 0.9216

Epoch 19/24
----------
train Loss: 0.2591 Acc: 0.9016
val Loss: 0.2489 Acc: 0.9216

Epoch 20/24
----------
train Loss: 0.2504 Acc: 0.8934
val Loss: 0.2442 Acc: 0.9216

Epoch 21/24
----------
train Loss: 0.2559 Acc: 0.8893
val Loss: 0.2181 Acc: 0.9281

Epoch 22/24
----------
train Loss: 0.3507 Acc: 0.8648
val Loss: 0.2282 Acc: 0.9281

Epoch 23/24
----------
train Loss: 0.2974 Acc: 0.8811
val Loss: 0.2046 Acc: 0.9412

Epoch 24/24
----------
train Loss: 0.3363 Acc: 0.8566
val Loss: 0.2129 Acc: 0.9346

Training complete in 1m 12s
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(weights='IMAGENET1K_V1')
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)

Epoch 0/24
----------
train Loss: 0.6298 Acc: 0.6598
val Loss: 0.2275 Acc: 0.9150

Epoch 1/24
----------
train Loss: 0.5977 Acc: 0.7500
val Loss: 0.1768 Acc: 0.9412

Epoch 2/24
----------
train Loss: 0.5647 Acc: 0.7418
val Loss: 0.1882 Acc: 0.9412

Epoch 3/24
----------
train Loss: 0.5057 Acc: 0.8074
val Loss: 0.4164 Acc: 0.8170

Epoch 4/24
----------
train Loss: 0.5201 Acc: 0.7541
val Loss: 0.2162 Acc: 0.9216

Epoch 5/24
----------
train Loss: 0.6274 Acc: 0.7623
val Loss: 0.6239 Acc: 0.7843

Epoch 6/24
----------
train Loss: 0.5192 Acc: 0.7828
val Loss: 0.2703 Acc: 0.9085

Epoch 7/24
----------
train Loss: 0.5037 Acc: 0.7869
val Loss: 0.1978 Acc: 0.9412

Epoch 8/24
----------
train Loss: 0.2841 Acc: 0.8730
val Loss: 0.1925 Acc: 0.9346

Epoch 9/24
----------
train Loss: 0.3668 Acc: 0.8320
val Loss: 0.2020 Acc: 0.9346

Epoch 10/24
----------
train Loss: 0.3743 Acc: 0.8361
val Loss: 0.2035 Acc: 0.9346

Epoch 11/24
----------
train Loss: 0.2629 Acc: 0.8648
val Loss: 0.2023 Acc: 0.9346

Epoch 12/24
----------
train Loss: 0.3484 Acc: 0.8607
val Loss: 0.1998 Acc: 0.9346

Epoch 13/24
----------
train Loss: 0.3568 Acc: 0.8525
val Loss: 0.1975 Acc: 0.9281

Epoch 14/24
----------
train Loss: 0.3894 Acc: 0.8238
val Loss: 0.2006 Acc: 0.9346

Epoch 15/24
----------
train Loss: 0.2698 Acc: 0.8811
val Loss: 0.1963 Acc: 0.9346

Epoch 16/24
----------
train Loss: 0.3503 Acc: 0.8607
val Loss: 0.2172 Acc: 0.9412

Epoch 17/24
----------
train Loss: 0.3145 Acc: 0.8811
val Loss: 0.1982 Acc: 0.9281

Epoch 18/24
----------
train Loss: 0.4361 Acc: 0.7910
val Loss: 0.2121 Acc: 0.9346

Epoch 19/24
----------
train Loss: 0.3733 Acc: 0.8566
val Loss: 0.1943 Acc: 0.9281

Epoch 20/24
----------
train Loss: 0.3832 Acc: 0.8320
val Loss: 0.2262 Acc: 0.9346

Epoch 21/24
----------
train Loss: 0.4323 Acc: 0.8443
val Loss: 0.2171 Acc: 0.9346

Epoch 22/24
----------
train Loss: 0.3286 Acc: 0.8689
val Loss: 0.2001 Acc: 0.9346

Epoch 23/24
----------
train Loss: 0.3509 Acc: 0.8320
val Loss: 0.2129 Acc: 0.9346

Epoch 24/24
----------
train Loss: 0.3919 Acc: 0.8484
val Loss: 0.2291 Acc: 0.9281

Training complete in 0m 47s
Best val Acc: 0.941176

visualize_model(model_conv)

plt.ioff()
plt.show()

Further Learning

If you would like to learn more about the applications of transfer learning, checkout our Quantized Transfer Learning for Computer Vision Tutorial.