Confusion matrix and test accuracy for PyTorch Transfer Learning tutorial

Question

Following the Pytorch Transfer learning tutorial, I am interested in reporting only train and test accuracy as well as confusion matrix (say using sklearn confusionmatrix). How can I do that? The current tutorial only reports train/val accuracy and I am having hard time figuring how to incorporate the sklearn confusionmatrix code there. Link to original tutorial here: https://pytorch.org/tutorials/beginner/transfer_learning_tutorial.html

%matplotlib inline
from graphviz import Digraph
import torch
from torch.autograd import Variable
# 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 numpy as np
import torchvision
from torchvision import datasets, models, transforms
import matplotlib.pyplot as plt
import time
import os
import copy

plt.ion()
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 = "images"
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")

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])

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('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()  # 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)

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

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('predicted: {}'.format(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)

model_ft = models.resnet18(pretrained=True)
num_ftrs = model_ft.fc.in_features
model_ft.fc = nn.Linear(num_ftrs, 9)

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)

model_ft = train_model(model_ft, criterion, optimizer_ft, exp_lr_scheduler,
                       num_epochs=25)

visualize_model(model_ft)

Solution 1

Answer given by ptrblck of PyTorch community. Thanks a lot!

nb_classes = 9

confusion_matrix = torch.zeros(nb_classes, nb_classes)
with torch.no_grad():
    for i, (inputs, classes) in enumerate(dataloaders['val']):
        inputs = inputs.to(device)
        classes = classes.to(device)
        outputs = model_ft(inputs)
        _, preds = torch.max(outputs, 1)
        for t, p in zip(classes.view(-1), preds.view(-1)):
                confusion_matrix[t.long(), p.long()] += 1

print(confusion_matrix)

To get the per-class accuracy:

print(confusion_matrix.diag()/confusion_matrix.sum(1))

Solution 2

Here is a slightly modified(direct) approach using sklearn's confusion_matrix:-

from sklearn.metrics import confusion_matrix

nb_classes = 9

# Initialize the prediction and label lists(tensors)
predlist=torch.zeros(0,dtype=torch.long, device='cpu')
lbllist=torch.zeros(0,dtype=torch.long, device='cpu')

with torch.no_grad():
    for i, (inputs, classes) in enumerate(dataloaders['val']):
        inputs = inputs.to(device)
        classes = classes.to(device)
        outputs = model_ft(inputs)
        _, preds = torch.max(outputs, 1)

        # Append batch prediction results
        predlist=torch.cat([predlist,preds.view(-1).cpu()])
        lbllist=torch.cat([lbllist,classes.view(-1).cpu()])

# Confusion matrix
conf_mat=confusion_matrix(lbllist.numpy(), predlist.numpy())
print(conf_mat)

# Per-class accuracy
class_accuracy=100*conf_mat.diagonal()/conf_mat.sum(1)
print(class_accuracy)

Solution 3

Follwing the answer above... Here is an answer with some visualization

nb_classes = 9
confusion_matrix = np.zeros((nb_classes, nb_classes))
with torch.no_grad():
    for i, (inputs, classes) in enumerate(test_loader):
        inputs = inputs.to(DEVICE)
        classes = classes.to(DEVICE)
        outputs = model(inputs)
        _, preds = torch.max(outputs, 1)
        for t, p in zip(classes.view(-1), preds.view(-1)):
                confusion_matrix[t.long(), p.long()] += 1

plt.figure(figsize=(15,10))

class_names = list(label2class.values())
df_cm = pd.DataFrame(confusion_matrix, index=class_names, columns=class_names).astype(int)
heatmap = sns.heatmap(df_cm, annot=True, fmt="d")

heatmap.yaxis.set_ticklabels(heatmap.yaxis.get_ticklabels(), rotation=0, ha='right',fontsize=15)
heatmap.xaxis.set_ticklabels(heatmap.xaxis.get_ticklabels(), rotation=45, ha='right',fontsize=15)
plt.ylabel('True label')
plt.xlabel('Predicted label')
;

Solution 4

I used the following to convert the torch tensors to an int defining the predicted class.

x = [torch.max(tensor).item() for tensor in x_data]
y = [torch.max(tensor).item() for tensor in y_data]

i hope this helps! i'm still a noob so please be gentle...

Solution 5

Another simple way to get accuracy is to use sklearns "accuracy_score". Heres an example:

from sklearn.metrics import accuracy_score
y_pred = y_pred.data.numpy()
accuracy = accuracy_score(labels, np.argmax(y_pred, axis=1))

First you need to get the data from the variable. "y_pred" is the predictions from your model, and labels are of course your labels.

np.argmax returns the index of the largest value inside the array. We want the largest value as it corresponds to the highest probability class when using softmax for multi-class classification. Accuracy score will return a percentage of matches between the labels and y_pred.

View more solutions

Comments

[talha06]

What is the dataloaders there? When I use my test_loader which is an instance of torch.utils.data.DataLoader class, get the TypeError: 'DataLoader' object is not subscriptable.

[donkey]

Why do you call t.long() and p.long()?

[Mustafa Aydın]

@donkey Because only integers, slices (`:`), ellipsis (`...`), None and long or byte Variables are valid indices

[Gulzar]

For a slot (row, col) - does it think row and predicted col, or the opposite? Please clarify

[Ozcan]

what is sns, and label2class?

[Melike]

Beware of nans with the above print.

[Sahar Millis]

seaborn, dictionary

[Admin]

@talha06 they seem to create a dataloaders object and divide it to train and val set. Here is an example from one of the Pytorch tutorials: dataloaders = {dl: DataLoader(ds, batch_size, shuffle=True) for dl, ds in (("train", train_ds), ("val", val_ds))}