Possible bug in SimpleITK or in PyTorch

SITK_PT.ipynb

{
  "nbformat": 4,
  "nbformat_minor": 0,
  "metadata": {
    "colab": {
      "name": "Untitled8.ipynb",
      "provenance": [],
      "authorship_tag": "ABX9TyMpidVlJ6mo6mAgV/r+FOI9",
      "include_colab_link": true
    },
    "kernelspec": {
      "name": "python3",
      "display_name": "Python 3"
    }
  },
  "cells": [
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "view-in-github",
        "colab_type": "text"
      },
      "source": [
        "<a href=\"https://colab.research.google.com/gist/fepegar/77a81c967c92a7c7f0a150cf0152940f/untitled8.ipynb\" target=\"_parent\"><img src=\"https://colab.research.google.com/assets/colab-badge.svg\" alt=\"Open In Colab\"/></a>"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "44q_UGrqziY4"
      },
      "source": [
        "!pip install --quiet "torchio==0.17.48"
      ],
      "execution_count": 1,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "inGkuZt4zlrc",
        "outputId": "20267eed-b40f-4874-aaed-18154c10aa58",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 105
        }
      },
      "source": [
        "import torch\n",
        "import torchio as tio\n",
        "import SimpleITK as sitk"
      ],
      "execution_count": 2,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "If you use TorchIO for your research, please cite the following paper:\n",
            "Pérez-García et al., TorchIO: a Python library for efficient loading,\n",
            "preprocessing, augmentation and patch-based sampling of medical images\n",
            "in deep learning. Credits instructions: https://torchio.readthedocs.io/#credits\n",
            "\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "cvfkmpK6zogM",
        "outputId": "ee56c9d2-8f14-403f-9ee7-a63502208815",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 72
        }
      },
      "source": [
        "transform = tio.RandomAffine()\n",
        "ixi = tio.datasets.IXITiny('ixi_tiny', download=True, transform=transform)\n",
        "\n",
        "paths = [ixi[i].image.path for i in range(4)]\n",
        "print(paths)\n",
        "\n",
        "class MyDataset(torch.utils.data.Dataset):\n",
        "    def __init__(self, paths):\n",
        "        self.paths = paths\n",
        "    \n",
        "    def __len__(self):\n",
        "        return len(self.paths)\n",
        "    \n",
        "    def __getitem__(self, index):\n",
        "        path = self.paths[index]\n",
        "        image = sitk.ReadImage(str(path))\n",
        "\n",
        "        resampler = sitk.ResampleImageFilter()\n",
        "        resampler.SetInterpolator(sitk.sitkNearestNeighbor)\n",
        "        resampler.SetReferenceImage(image)\n",
        "        resampler.SetDefaultPixelValue(0.0)\n",
        "        resampler.SetOutputPixelType(sitk.sitkFloat32)\n",
        "        resampled = resampler.Execute(image)\n",
        "\n",
        "        return sitk.GetArrayFromImage(resampled)\n",
        "\n",
        "my_dataset = MyDataset(paths)\n",
        "\n",
        "loader_sp = torch.utils.data.DataLoader(my_dataset, batch_size=4, num_workers=0)\n",
        "loader_mp = torch.utils.data.DataLoader(my_dataset, batch_size=4, num_workers=2)"
      ],
      "execution_count": 3,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Root directory for IXITiny found: ixi_tiny\n",
            "[PosixPath('ixi_tiny/image/IXI002-Guys-0828_image.nii.gz'), PosixPath('ixi_tiny/image/IXI012-HH-1211_image.nii.gz'), PosixPath('ixi_tiny/image/IXI013-HH-1212_image.nii.gz'), PosixPath('ixi_tiny/image/IXI014-HH-1236_image.nii.gz')]\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "4LZ1MUJb0CVn",
        "outputId": "68bb29a5-53bb-4194-a12a-aedd7ecd7c9e",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 34
        }
      },
      "source": [
        "batch_sp = next(iter(loader_sp))\n",
        "print(batch_sp.shape)"
      ],
      "execution_count": 4,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "torch.Size([4, 55, 44, 83])\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "WEckhVxE0KHQ"
      },
      "source": [
        "batch_mp = next(iter(loader_mp))\n",
        "print(batch_mp.shape)"
      ],
      "execution_count": null,
      "outputs": []
    }
  ]
}

The code, just in case:

# -*- coding: utf-8 -*-
"""Untitled8.ipynb

Automatically generated by Colaboratory.

Original file is located at
    https://colab.research.google.com/drive/1GzO2RQYn7jI26frWS8S-YlbSmjFAehWa
"""

# pip install --quiet torchio==0.17.48

import torch
import torchio as tio
import SimpleITK as sitk

transform = tio.RandomAffine()
ixi = tio.datasets.IXITiny('ixi_tiny', download=True, transform=transform)

paths = [ixi[i].image.path for i in range(4)]
print(paths)


class MyDataset(torch.utils.data.Dataset):
    def __init__(self, paths):
        self.paths = paths

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

    def __getitem__(self, index):
        path = self.paths[index]
        image = sitk.ReadImage(str(path))

        resampler = sitk.ResampleImageFilter()
        resampler.SetInterpolator(sitk.sitkNearestNeighbor)
        resampler.SetReferenceImage(image)
        resampler.SetDefaultPixelValue(0.0)
        resampler.SetOutputPixelType(sitk.sitkFloat32)
        resampled = resampler.Execute(image)

        return sitk.GetArrayFromImage(resampled)

my_dataset = MyDataset(paths)

loader_sp = torch.utils.data.DataLoader(my_dataset, batch_size=4, num_workers=0)
loader_mp = torch.utils.data.DataLoader(my_dataset, batch_size=4, num_workers=2)

batch_sp = next(iter(loader_sp))
print(batch_sp.shape)

batch_mp = next(iter(loader_mp))
print(batch_mp.shape)