Simple sequential MNIST dataset implementation in Python and PyTorch for usage in recurrent neural networks. See for project using this dataset here: https://github.com/V0XNIHILI/LSTM-Sequential-MNIST

example_usage.py

import torchvision
import torchvision.transforms as transforms

MNIST_MEAN = 0.1307
MNIST_STD = 0.3081

mnist_transforms = transforms.Compose([transforms.ToTensor(), transforms.Normalize((MNIST_MEAN,), (MNIST_STD,))])

train_data = SequentialMNIST(torchvision.datasets.MNIST('.', train=True, download=True, transform=mnist_transforms))

train_data_loader = DataLoader(train_data, batch_size=64)

SequentialMNIST.py

import torch
import torch.nn as nn
import torchvision
from torch.utils.data import Dataset

class SequentialMNIST(Dataset):

    def __init__(self, MNIST_dataset: torchvision.datasets.MNIST):
        self.MNIST_dataset = MNIST_dataset

    def __len__(self):
        return len(self.MNIST_dataset)

    def __getitem__(self, idx: int):
        image, target = self.MNIST_dataset[idx]

        # Return flattened image in shape (time steps = 784, dimensionality = 1)
        return image.view(-1, 1), target