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
import copy

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

<contextlib.ExitStack object at 0x7f4cbd5cee90>

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(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
                best_model_wts = copy.deepcopy(model.state_dict())

        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(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(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(pretrained=True)
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)

/opt/conda/lib/python3.10/site-packages/torchvision/models/_utils.py:208: UserWarning:

The parameter 'pretrained' is deprecated since 0.13 and may be removed in the future, please use 'weights' instead.

/opt/conda/lib/python3.10/site-packages/torchvision/models/_utils.py:223: UserWarning:

Arguments other than a weight enum or `None` for 'weights' are deprecated since 0.13 and may be removed in the future. The current behavior is equivalent to passing `weights=ResNet18_Weights.IMAGENET1K_V1`. You can also use `weights=ResNet18_Weights.DEFAULT` to get the most up-to-date weights.

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]
 27%|##6       | 12.0M/44.7M [00:00<00:00, 125MB/s]
 87%|########7 | 38.9M/44.7M [00:00<00:00, 218MB/s]
100%|##########| 44.7M/44.7M [00:00<00:00, 211MB/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: 0.5014 Acc: 0.7541
val Loss: 0.2609 Acc: 0.9216

Epoch 1/24
----------
train Loss: 0.4872 Acc: 0.7869
val Loss: 0.3096 Acc: 0.8889

Epoch 2/24
----------
train Loss: 0.4673 Acc: 0.7951
val Loss: 0.3405 Acc: 0.8693

Epoch 3/24
----------
train Loss: 0.5813 Acc: 0.7664
val Loss: 0.5761 Acc: 0.8301

Epoch 4/24
----------
train Loss: 0.7486 Acc: 0.7541
val Loss: 0.4635 Acc: 0.8497

Epoch 5/24
----------
train Loss: 0.4865 Acc: 0.8238
val Loss: 0.4161 Acc: 0.8954

Epoch 6/24
----------
train Loss: 0.9054 Acc: 0.7377
val Loss: 0.5279 Acc: 0.8235

Epoch 7/24
----------
train Loss: 0.9062 Acc: 0.7418
val Loss: 0.4232 Acc: 0.9020

Epoch 8/24
----------
train Loss: 0.4454 Acc: 0.8730
val Loss: 0.3637 Acc: 0.9020

Epoch 9/24
----------
train Loss: 0.3090 Acc: 0.8648
val Loss: 0.3067 Acc: 0.9085

Epoch 10/24
----------
train Loss: 0.3434 Acc: 0.8648
val Loss: 0.2951 Acc: 0.9150

Epoch 11/24
----------
train Loss: 0.3824 Acc: 0.8402
val Loss: 0.2968 Acc: 0.9281

Epoch 12/24
----------
train Loss: 0.3464 Acc: 0.8607
val Loss: 0.3931 Acc: 0.8824

Epoch 13/24
----------
train Loss: 0.2935 Acc: 0.8811
val Loss: 0.2677 Acc: 0.9085

Epoch 14/24
----------
train Loss: 0.2395 Acc: 0.9098
val Loss: 0.2591 Acc: 0.9020

Epoch 15/24
----------
train Loss: 0.2440 Acc: 0.8893
val Loss: 0.3148 Acc: 0.9020

Epoch 16/24
----------
train Loss: 0.3113 Acc: 0.8811
val Loss: 0.2714 Acc: 0.9150

Epoch 17/24
----------
train Loss: 0.2631 Acc: 0.8975
val Loss: 0.2448 Acc: 0.9281

Epoch 18/24
----------
train Loss: 0.3016 Acc: 0.8852
val Loss: 0.3498 Acc: 0.8954

Epoch 19/24
----------
train Loss: 0.3080 Acc: 0.8893
val Loss: 0.2658 Acc: 0.9085

Epoch 20/24
----------
train Loss: 0.2357 Acc: 0.8975
val Loss: 0.2600 Acc: 0.9150

Epoch 21/24
----------
train Loss: 0.2691 Acc: 0.8975
val Loss: 0.2756 Acc: 0.9020

Epoch 22/24
----------
train Loss: 0.2439 Acc: 0.8852
val Loss: 0.2704 Acc: 0.9150

Epoch 23/24
----------
train Loss: 0.2286 Acc: 0.8934
val Loss: 0.2672 Acc: 0.9150

Epoch 24/24
----------
train Loss: 0.2712 Acc: 0.9098
val Loss: 0.3618 Acc: 0.8954

Training complete in 1m 15s
Best val Acc: 0.928105

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)

/opt/conda/lib/python3.10/site-packages/torchvision/models/_utils.py:208: UserWarning:

The parameter 'pretrained' is deprecated since 0.13 and may be removed in the future, please use 'weights' instead.

/opt/conda/lib/python3.10/site-packages/torchvision/models/_utils.py:223: UserWarning:

Arguments other than a weight enum or `None` for 'weights' are deprecated since 0.13 and may be removed in the future. The current behavior is equivalent to passing `weights=ResNet18_Weights.IMAGENET1K_V1`. You can also use `weights=ResNet18_Weights.DEFAULT` to get the most up-to-date weights.

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: 0.5699 Acc: 0.6803
val Loss: 0.2760 Acc: 0.8824

Epoch 1/24
----------
train Loss: 0.4527 Acc: 0.7746
val Loss: 0.1992 Acc: 0.9412

Epoch 2/24
----------
train Loss: 0.4561 Acc: 0.7951
val Loss: 0.2337 Acc: 0.9216

Epoch 3/24
----------
train Loss: 0.7419 Acc: 0.7008
val Loss: 0.1953 Acc: 0.9412

Epoch 4/24
----------
train Loss: 0.4166 Acc: 0.8320
val Loss: 0.1743 Acc: 0.9477

Epoch 5/24
----------
train Loss: 0.4538 Acc: 0.8074
val Loss: 0.1847 Acc: 0.9542

Epoch 6/24
----------
train Loss: 0.5023 Acc: 0.8033
val Loss: 0.1940 Acc: 0.9477

Epoch 7/24
----------
train Loss: 0.4706 Acc: 0.8238
val Loss: 0.1736 Acc: 0.9412

Epoch 8/24
----------
train Loss: 0.3064 Acc: 0.8607
val Loss: 0.1934 Acc: 0.9412

Epoch 9/24
----------
train Loss: 0.3855 Acc: 0.8402
val Loss: 0.1926 Acc: 0.9412

Epoch 10/24
----------
train Loss: 0.3858 Acc: 0.8443
val Loss: 0.1946 Acc: 0.9477

Epoch 11/24
----------
train Loss: 0.3845 Acc: 0.8197
val Loss: 0.1978 Acc: 0.9412

Epoch 12/24
----------
train Loss: 0.5324 Acc: 0.7787
val Loss: 0.1938 Acc: 0.9542

Epoch 13/24
----------
train Loss: 0.3325 Acc: 0.8648
val Loss: 0.1894 Acc: 0.9477

Epoch 14/24
----------
train Loss: 0.2851 Acc: 0.8811
val Loss: 0.2073 Acc: 0.9412

Epoch 15/24
----------
train Loss: 0.2335 Acc: 0.9057
val Loss: 0.1801 Acc: 0.9281

Epoch 16/24
----------
train Loss: 0.3294 Acc: 0.8648
val Loss: 0.1806 Acc: 0.9477

Epoch 17/24
----------
train Loss: 0.3842 Acc: 0.8238
val Loss: 0.1861 Acc: 0.9477

Epoch 18/24
----------
train Loss: 0.3214 Acc: 0.8484
val Loss: 0.2688 Acc: 0.9085

Epoch 19/24
----------
train Loss: 0.3105 Acc: 0.8361
val Loss: 0.2020 Acc: 0.9412

Epoch 20/24
----------
train Loss: 0.3614 Acc: 0.8607
val Loss: 0.1794 Acc: 0.9477

Epoch 21/24
----------
train Loss: 0.3183 Acc: 0.8361
val Loss: 0.2188 Acc: 0.9412

Epoch 22/24
----------
train Loss: 0.3029 Acc: 0.8648
val Loss: 0.1947 Acc: 0.9412

Epoch 23/24
----------
train Loss: 0.3476 Acc: 0.8525
val Loss: 0.1756 Acc: 0.9477

Epoch 24/24
----------
train Loss: 0.3963 Acc: 0.8197
val Loss: 0.2043 Acc: 0.9477

Training complete in 0m 45s
Best val Acc: 0.954248

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.