Transfer Learning for Computer Vision Tutorial

Created On: Mar 24, 2017 | Last Updated: Jan 27, 2025 | Last Verified: Nov 05, 2024

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

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 PIL import Image
from tempfile import TemporaryDirectory

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

<contextlib.ExitStack object at 0x7f4d8a2a0dc0>

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

# We want to be able to train our model on an `accelerator <https://pytorch.org/docs/stable/torch.html#accelerators>`__
# such as CUDA, MPS, MTIA, or XPU. If the current accelerator is available, we will use it. Otherwise, we use the CPU.

device = torch.accelerator.current_accelerator().type if torch.accelerator.is_available() else "cpu"
print(f"Using {device} device")

Using cuda device

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, weights_only=True))
    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/ci-user/.cache/torch/hub/checkpoints/resnet18-f37072fd.pth

  0%|          | 0.00/44.7M [00:00<?, ?B/s]
 92%|#########1| 40.9M/44.7M [00:00<00:00, 427MB/s]
100%|##########| 44.7M/44.7M [00:00<00:00, 429MB/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: 0.5934 Acc: 0.6762
val Loss: 0.1839 Acc: 0.9477

Epoch 1/24
----------
train Loss: 0.5242 Acc: 0.7459
val Loss: 0.3569 Acc: 0.8497

Epoch 2/24
----------
train Loss: 0.5851 Acc: 0.7623
val Loss: 0.5686 Acc: 0.8235

Epoch 3/24
----------
train Loss: 0.5045 Acc: 0.7992
val Loss: 0.3031 Acc: 0.8627

Epoch 4/24
----------
train Loss: 0.3849 Acc: 0.8484
val Loss: 0.4061 Acc: 0.8497

Epoch 5/24
----------
train Loss: 0.4241 Acc: 0.8238
val Loss: 0.2893 Acc: 0.8824

Epoch 6/24
----------
train Loss: 0.4898 Acc: 0.8156
val Loss: 0.2659 Acc: 0.9085

Epoch 7/24
----------
train Loss: 0.3531 Acc: 0.8402
val Loss: 0.1874 Acc: 0.9281

Epoch 8/24
----------
train Loss: 0.3061 Acc: 0.8648
val Loss: 0.2018 Acc: 0.9281

Epoch 9/24
----------
train Loss: 0.3024 Acc: 0.8770
val Loss: 0.1760 Acc: 0.9281

Epoch 10/24
----------
train Loss: 0.3213 Acc: 0.8689
val Loss: 0.1837 Acc: 0.9216

Epoch 11/24
----------
train Loss: 0.1981 Acc: 0.9303
val Loss: 0.1948 Acc: 0.9216

Epoch 12/24
----------
train Loss: 0.3077 Acc: 0.8689
val Loss: 0.1812 Acc: 0.9150

Epoch 13/24
----------
train Loss: 0.2675 Acc: 0.8975
val Loss: 0.1687 Acc: 0.9281

Epoch 14/24
----------
train Loss: 0.2867 Acc: 0.8689
val Loss: 0.1859 Acc: 0.9346

Epoch 15/24
----------
train Loss: 0.2691 Acc: 0.8811
val Loss: 0.1795 Acc: 0.9281

Epoch 16/24
----------
train Loss: 0.2679 Acc: 0.8975
val Loss: 0.1678 Acc: 0.9281

Epoch 17/24
----------
train Loss: 0.2936 Acc: 0.8811
val Loss: 0.1742 Acc: 0.9412

Epoch 18/24
----------
train Loss: 0.2228 Acc: 0.8893
val Loss: 0.1600 Acc: 0.9281

Epoch 19/24
----------
train Loss: 0.2757 Acc: 0.8852
val Loss: 0.1935 Acc: 0.9216

Epoch 20/24
----------
train Loss: 0.2405 Acc: 0.9057
val Loss: 0.1886 Acc: 0.9085

Epoch 21/24
----------
train Loss: 0.2574 Acc: 0.8934
val Loss: 0.1783 Acc: 0.9281

Epoch 22/24
----------
train Loss: 0.2851 Acc: 0.8770
val Loss: 0.1574 Acc: 0.9216

Epoch 23/24
----------
train Loss: 0.2819 Acc: 0.8975
val Loss: 0.1588 Acc: 0.9346

Epoch 24/24
----------
train Loss: 0.2808 Acc: 0.8893
val Loss: 0.2257 Acc: 0.9150

Training complete in 0m 33s
Best val Acc: 0.947712

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: 0.5237 Acc: 0.7131
val Loss: 0.1926 Acc: 0.9412

Epoch 1/24
----------
train Loss: 0.4499 Acc: 0.8115
val Loss: 0.2324 Acc: 0.9216

Epoch 2/24
----------
train Loss: 0.5221 Acc: 0.7705
val Loss: 0.1761 Acc: 0.9477

Epoch 3/24
----------
train Loss: 0.5662 Acc: 0.7705
val Loss: 0.2574 Acc: 0.9020

Epoch 4/24
----------
train Loss: 0.5073 Acc: 0.7787
val Loss: 0.1704 Acc: 0.9477

Epoch 5/24
----------
train Loss: 0.5519 Acc: 0.7746
val Loss: 0.2246 Acc: 0.8954

Epoch 6/24
----------
train Loss: 0.5828 Acc: 0.7828
val Loss: 0.2307 Acc: 0.9150

Epoch 7/24
----------
train Loss: 0.3763 Acc: 0.8484
val Loss: 0.1722 Acc: 0.9542

Epoch 8/24
----------
train Loss: 0.3300 Acc: 0.8607
val Loss: 0.1708 Acc: 0.9477

Epoch 9/24
----------
train Loss: 0.3177 Acc: 0.8484
val Loss: 0.1600 Acc: 0.9542

Epoch 10/24
----------
train Loss: 0.2990 Acc: 0.8730
val Loss: 0.1786 Acc: 0.9477

Epoch 11/24
----------
train Loss: 0.4442 Acc: 0.7951
val Loss: 0.1728 Acc: 0.9608

Epoch 12/24
----------
train Loss: 0.3718 Acc: 0.8238
val Loss: 0.1712 Acc: 0.9542

Epoch 13/24
----------
train Loss: 0.3371 Acc: 0.8566
val Loss: 0.1604 Acc: 0.9542

Epoch 14/24
----------
train Loss: 0.3018 Acc: 0.8566
val Loss: 0.1727 Acc: 0.9542

Epoch 15/24
----------
train Loss: 0.3567 Acc: 0.8238
val Loss: 0.1549 Acc: 0.9542

Epoch 16/24
----------
train Loss: 0.3718 Acc: 0.8279
val Loss: 0.1840 Acc: 0.9412

Epoch 17/24
----------
train Loss: 0.2899 Acc: 0.8648
val Loss: 0.1891 Acc: 0.9281

Epoch 18/24
----------
train Loss: 0.3534 Acc: 0.8443
val Loss: 0.1679 Acc: 0.9542

Epoch 19/24
----------
train Loss: 0.3147 Acc: 0.8484
val Loss: 0.1687 Acc: 0.9542

Epoch 20/24
----------
train Loss: 0.3169 Acc: 0.8648
val Loss: 0.2108 Acc: 0.9281

Epoch 21/24
----------
train Loss: 0.3107 Acc: 0.8689
val Loss: 0.1713 Acc: 0.9542

Epoch 22/24
----------
train Loss: 0.3286 Acc: 0.8197
val Loss: 0.1702 Acc: 0.9542

Epoch 23/24
----------
train Loss: 0.3996 Acc: 0.8115
val Loss: 0.1674 Acc: 0.9542

Epoch 24/24
----------
train Loss: 0.3337 Acc: 0.8648
val Loss: 0.1681 Acc: 0.9542

Training complete in 0m 27s
Best val Acc: 0.960784

visualize_model(model_conv)

plt.ioff()
plt.show()

Inference on custom images

Use the trained model to make predictions on custom images and visualize the predicted class labels along with the images.

def visualize_model_predictions(model,img_path):
    was_training = model.training
    model.eval()

    img = Image.open(img_path)
    img = data_transforms['val'](img)
    img = img.unsqueeze(0)
    img = img.to(device)

    with torch.no_grad():
        outputs = model(img)
        _, preds = torch.max(outputs, 1)

        ax = plt.subplot(2,2,1)
        ax.axis('off')
        ax.set_title(f'Predicted: {class_names[preds[0]]}')
        imshow(img.cpu().data[0])

        model.train(mode=was_training)

visualize_model_predictions(
    model_conv,
    img_path='data/hymenoptera_data/val/bees/72100438_73de9f17af.jpg'
)

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.