Given input size: (256x1x1). Calculated output size: (256x0x0)

Hello,

I have this issue: Given input size: (256x1x1). Calculated output size: (256x0x0)
I found some similar problems on google, but could not find my solution.

I use the CIFAR10 dataset (32x32 pictures). Is it the issue? Should I change the dimensions of it?
As I use the alexnet model from Pytorch, I don’t see any other reasons why I should have a problem.

Thanks for your help
Below is my code:

import torch
import torch.nn as nn
import torch.nn.functional as F
import torch
import torch.optim as optim
import torch.nn.functional as F
import torch.backends.cudnn as cudnn
import torchvision.transforms as transforms
import torchvision

import argparse

path = "/Users/.../data"



parser = argparse.ArgumentParser(description='PyTorch CIFAR10 Training')
parser.add_argument('--lr', default=0.01, type=float, help='learning rate')
parser.add_argument('--resume', '-r', action='store_true', help='resume from checkpoint')
args = parser.parse_args()  



device = 'cuda' if torch.cuda.is_available() else 'cpu'
best_acc = 0  # best test accuracy
start_epoch = 0  # start from epoch 0 or last checkpoint epoch

# Data
print('==> Preparing data..')
transform_train = transforms.Compose([
    transforms.RandomCrop(32, padding=4),
    transforms.RandomHorizontalFlip(),
    transforms.ToTensor(),
    transforms.Normalize((0.4914, 0.4822, 0.4465), (0.2023, 0.1994, 0.2010)),
])

transform_test = transforms.Compose([
    transforms.ToTensor(),
    transforms.Normalize((0.4914, 0.4822, 0.4465), (0.2023, 0.1994, 0.2010)),
])

from torchvision.datasets import CIFAR10
training_set = CIFAR10(path,train=True, transform=transform_train)
#trainset = torchvision.datasets.CIFAR10(root='./data', train=True, download=True, transform=transform_train)
trainset = training_set
trainloader = torch.utils.data.DataLoader(trainset, batch_size=10, shuffle=True, num_workers=2)


testset = CIFAR10(path, train=False, download=True, transform=transform_test)
testloader = torch.utils.data.DataLoader(testset, batch_size=10, shuffle=False, num_workers=2)

classes = ('plane', 'car', 'bird', 'cat', 'deer', 'dog', 'frog', 'horse', 'ship', 'truck')

# Model
print('==> Building model..')

import torch.utils
import torchvision.models as models
alexnet = models.alexnet(pretrained=True)

# Freeze model weights
for param in alexnet.parameters():
    param.requires_grad = False
print(alexnet)

# Modify classifier
alexnet.classifier[6] = nn.Linear(4096, 10)                       
print(alexnet)
net = alexnet

if device == 'cuda':
    net = torch.nn.DataParallel(net)
    cudnn.benchmark = True

if args.resume:
    # Load checkpoint.
    print('==> Resuming from checkpoint..')
    assert os.path.isdir('checkpoint'), 'Error: no checkpoint directory found!'
    checkpoint = torch.load('./checkpoint/ckpt.t7')
    net.load_state_dict(checkpoint['net'])
    best_acc = checkpoint['acc']
    start_epoch = checkpoint['epoch']

criterion = nn.CrossEntropyLoss()
optimizer = optim.SGD(net.parameters(), lr=args.lr, momentum=0.9, weight_decay=5e-4)

# Training
def train(epoch):
    print('\nEpoch: %d' % epoch)
    net.train()
    train_loss = 0
    correct = 0
    total = 0
    for batch_idx, (inputs, targets) in enumerate(trainloader):
        inputs, targets = inputs.to(device), targets.to(device)
        optimizer.zero_grad()
        outputs = net(inputs)
        loss = criterion(outputs, targets)
        loss.backward()
        optimizer.step()

        train_loss += loss.item()
        _, predicted = outputs.max(1)
        total += targets.size(0)
        correct += predicted.eq(targets).sum().item()

        progress_bar(batch_idx, len(trainloader), 'Loss: %.3f | Acc: %.3f%% (%d/%d)'
            % (train_loss/(batch_idx+1), 100.*correct/total, correct, total))

def test(epoch):
    global best_acc
    net.eval()
    test_loss = 0
    correct = 0
    total = 0
    with torch.no_grad():
        for batch_idx, (inputs, targets) in enumerate(testloader):
            inputs, targets = inputs.to(device), targets.to(device)
            outputs = net(inputs)
            loss = criterion(outputs, targets)

            test_loss += loss.item()
            _, predicted = outputs.max(1)
            total += targets.size(0)
            correct += predicted.eq(targets).sum().item()

            progress_bar(batch_idx, len(testloader), 'Loss: %.3f | Acc: %.3f%% (%d/%d)'
                % (test_loss/(batch_idx+1), 100.*correct/total, correct, total))

    # Save checkpoint.
    acc = 100.*correct/total
    if acc > best_acc:
        print('Saving..')
        state = {
            'net': net.state_dict(),
            'acc': acc,
            'epoch': epoch,
        }
        if not os.path.isdir('checkpoint'):
            os.mkdir('checkpoint')
        torch.save(state, './checkpoint/ckpt.t7')
        best_acc = acc


for epoch in range(start_epoch, start_epoch+10):
    train(epoch)

test(epoch)


The spatial size of your input images is too small for this model.
You could try to resize the images to the original input size (224x224) or create a custom model suitable for these image size.

Wow, that’s all?

And how can I properly resize my image?

Or, maybe, is there an other pytorch architecture with a pretrained parameter that would better fit this picture size?

The Training a Classifier tutorial uses the CIFAR10 dataset, so you might take a look at their model and use it as a baseline.

If you would like to resize your images, you could use:

transform_train = transforms.Compose([
    transforms.Resize(224),
    transforms.RandomHorizontalFlip(),
    transforms.ToTensor(),
    transforms.Normalize((0.4914, 0.4822, 0.4465), (0.2023, 0.1994, 0.2010)),
])

transform_test = transforms.Compose([
    transforms.Resize(224),
    transforms.ToTensor(),
    transforms.Normalize((0.4914, 0.4822, 0.4465), (0.2023, 0.1994, 0.2010)),
])

If you want to use the random cropping, you should resize the images to a larger size (e.g. 240) and crop them to 224.