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 0x7f68bb2b3df0>

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()

    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(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]
 15%|#5        | 6.80M/44.7M [00:00<00:00, 71.1MB/s]
 37%|###6      | 16.3M/44.7M [00:00<00:00, 87.9MB/s]
 80%|########  | 35.8M/44.7M [00:00<00:00, 141MB/s]
100%|##########| 44.7M/44.7M [00:00<00:00, 136MB/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: 0.5788 Acc: 0.7131
val Loss: 0.2115 Acc: 0.9216

Epoch 1/24
----------
train Loss: 0.4469 Acc: 0.7869
val Loss: 0.3251 Acc: 0.8824

Epoch 2/24
----------
train Loss: 0.4572 Acc: 0.8115
val Loss: 0.1631 Acc: 0.9216

Epoch 3/24
----------
train Loss: 0.5114 Acc: 0.7910
val Loss: 0.3916 Acc: 0.9020

Epoch 4/24
----------
train Loss: 0.3799 Acc: 0.8443
val Loss: 0.2946 Acc: 0.8889

Epoch 5/24
----------
train Loss: 0.6323 Acc: 0.7746
val Loss: 0.2516 Acc: 0.8889

Epoch 6/24
----------
train Loss: 0.5754 Acc: 0.7787
val Loss: 0.2751 Acc: 0.8954

Epoch 7/24
----------
train Loss: 0.3900 Acc: 0.8320
val Loss: 0.2252 Acc: 0.9216

Epoch 8/24
----------
train Loss: 0.3147 Acc: 0.8770
val Loss: 0.2324 Acc: 0.9085

Epoch 9/24
----------
train Loss: 0.2700 Acc: 0.8730
val Loss: 0.2509 Acc: 0.9085

Epoch 10/24
----------
train Loss: 0.3676 Acc: 0.8566
val Loss: 0.2271 Acc: 0.9085

Epoch 11/24
----------
train Loss: 0.2465 Acc: 0.8893
val Loss: 0.2318 Acc: 0.9085

Epoch 12/24
----------
train Loss: 0.3220 Acc: 0.8689
val Loss: 0.1984 Acc: 0.9216

Epoch 13/24
----------
train Loss: 0.2460 Acc: 0.8934
val Loss: 0.2595 Acc: 0.8954

Epoch 14/24
----------
train Loss: 0.1964 Acc: 0.9262
val Loss: 0.2266 Acc: 0.9150

Epoch 15/24
----------
train Loss: 0.3581 Acc: 0.8648
val Loss: 0.2201 Acc: 0.9216

Epoch 16/24
----------
train Loss: 0.2126 Acc: 0.9057
val Loss: 0.2255 Acc: 0.9085

Epoch 17/24
----------
train Loss: 0.2741 Acc: 0.8934
val Loss: 0.2315 Acc: 0.9085

Epoch 18/24
----------
train Loss: 0.2914 Acc: 0.8893
val Loss: 0.2383 Acc: 0.9020

Epoch 19/24
----------
train Loss: 0.2430 Acc: 0.8770
val Loss: 0.2077 Acc: 0.9020

Epoch 20/24
----------
train Loss: 0.3027 Acc: 0.8893
val Loss: 0.2234 Acc: 0.9085

Epoch 21/24
----------
train Loss: 0.2725 Acc: 0.8852
val Loss: 0.2363 Acc: 0.9020

Epoch 22/24
----------
train Loss: 0.1699 Acc: 0.9385
val Loss: 0.2191 Acc: 0.9216

Epoch 23/24
----------
train Loss: 0.2241 Acc: 0.9016
val Loss: 0.2057 Acc: 0.9150

Epoch 24/24
----------
train Loss: 0.2481 Acc: 0.8893
val Loss: 0.2265 Acc: 0.9085

Training complete in 1m 10s
Best val Acc: 0.921569

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: 0.6403 Acc: 0.6311
val Loss: 0.2682 Acc: 0.8889

Epoch 1/24
----------
train Loss: 0.4362 Acc: 0.7951
val Loss: 0.2471 Acc: 0.9150

Epoch 2/24
----------
train Loss: 0.4381 Acc: 0.8279
val Loss: 0.4092 Acc: 0.8431

Epoch 3/24
----------
train Loss: 0.5624 Acc: 0.7705
val Loss: 0.2051 Acc: 0.9281

Epoch 4/24
----------
train Loss: 0.4282 Acc: 0.8115
val Loss: 0.2408 Acc: 0.9085

Epoch 5/24
----------
train Loss: 0.4510 Acc: 0.7828
val Loss: 0.3151 Acc: 0.8824

Epoch 6/24
----------
train Loss: 0.4558 Acc: 0.8156
val Loss: 0.2153 Acc: 0.9150

Epoch 7/24
----------
train Loss: 0.3573 Acc: 0.8525
val Loss: 0.2112 Acc: 0.9216

Epoch 8/24
----------
train Loss: 0.3806 Acc: 0.8115
val Loss: 0.1669 Acc: 0.9281

Epoch 9/24
----------
train Loss: 0.3773 Acc: 0.8238
val Loss: 0.1744 Acc: 0.9412

Epoch 10/24
----------
train Loss: 0.2814 Acc: 0.8607
val Loss: 0.2322 Acc: 0.9020

Epoch 11/24
----------
train Loss: 0.3314 Acc: 0.8648
val Loss: 0.2071 Acc: 0.9281

Epoch 12/24
----------
train Loss: 0.2969 Acc: 0.8730
val Loss: 0.2010 Acc: 0.9346

Epoch 13/24
----------
train Loss: 0.3091 Acc: 0.8811
val Loss: 0.1765 Acc: 0.9412

Epoch 14/24
----------
train Loss: 0.3220 Acc: 0.8566
val Loss: 0.1891 Acc: 0.9412

Epoch 15/24
----------
train Loss: 0.3223 Acc: 0.8402
val Loss: 0.1773 Acc: 0.9412

Epoch 16/24
----------
train Loss: 0.3788 Acc: 0.8525
val Loss: 0.2019 Acc: 0.9216

Epoch 17/24
----------
train Loss: 0.2785 Acc: 0.8607
val Loss: 0.1929 Acc: 0.9281

Epoch 18/24
----------
train Loss: 0.2819 Acc: 0.8689
val Loss: 0.1753 Acc: 0.9346

Epoch 19/24
----------
train Loss: 0.3174 Acc: 0.8566
val Loss: 0.1904 Acc: 0.9346

Epoch 20/24
----------
train Loss: 0.2524 Acc: 0.8811
val Loss: 0.1822 Acc: 0.9412

Epoch 21/24
----------
train Loss: 0.2663 Acc: 0.8689
val Loss: 0.1995 Acc: 0.9346

Epoch 22/24
----------
train Loss: 0.3594 Acc: 0.8402
val Loss: 0.1888 Acc: 0.9412

Epoch 23/24
----------
train Loss: 0.3443 Acc: 0.8484
val Loss: 0.1833 Acc: 0.9346

Epoch 24/24
----------
train Loss: 0.2670 Acc: 0.8811
val Loss: 0.1721 Acc: 0.9346

Training complete in 0m 43s
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.