.. DO NOT EDIT.
.. THIS FILE WAS AUTOMATICALLY GENERATED BY SPHINX-GALLERY.
.. TO MAKE CHANGES, EDIT THE SOURCE PYTHON FILE:
.. "auto_examples/detectors/mmahalanobis.py"
.. LINE NUMBERS ARE GIVEN BELOW.

.. only:: html

    .. note::
        :class: sphx-glr-download-link-note

        :ref:`Go to the end <sphx_glr_download_auto_examples_detectors_mmahalanobis.py>`
        to download the full example code.

.. rst-class:: sphx-glr-example-title

.. _sphx_glr_auto_examples_detectors_mmahalanobis.py:


Multi-Layer Mahalanobis
==============================

Running :class:`MultiMahalanobis <pytorch_ood.detector.MultiMahalanobis>` on CIFAR 10.

.. GENERATED FROM PYTHON SOURCE LINES 8-26

.. code-block:: Python
   :lineno-start: 9


    import logging

    from torch import nn
    from torch.utils.data import DataLoader
    from torchvision.datasets import CIFAR10

    from pytorch_ood.dataset.img import Textures
    from pytorch_ood.detector import MultiMahalanobis
    from pytorch_ood.model import WideResNet
    from pytorch_ood.utils import OODMetrics, ToUnknown, fix_random_seed

    logging.basicConfig(level=logging.INFO)

    fix_random_seed(123)

    device = "cuda"


.. GENERATED FROM PYTHON SOURCE LINES 27-28

Setup preprocessing and data

.. GENERATED FROM PYTHON SOURCE LINES 28-41

.. code-block:: Python
   :lineno-start: 28

    trans = WideResNet.transform_for("cifar10-pt")

    dataset_train = CIFAR10(root="data", train=True, download=True, transform=trans)
    dataset_in_test = CIFAR10(root="data", train=False, download=True, transform=trans)
    dataset_out_test = Textures(
        root="data", download=True, transform=trans, target_transform=ToUnknown()
    )

    train_loader = DataLoader(dataset_train, batch_size=128, shuffle=True)

    # create data loaders
    test_loader = DataLoader(dataset_in_test + dataset_out_test, batch_size=128)


.. GENERATED FROM PYTHON SOURCE LINES 42-43

Stage 1: Create DNN pre-trained on CIFAR 10

.. GENERATED FROM PYTHON SOURCE LINES 43-51

.. code-block:: Python
   :lineno-start: 43

    model = WideResNet(num_classes=10, pretrained="cifar10-pt").to(device).eval()

    layer1 = model.conv1
    layer2 = model.block1
    layer3 = model.block2
    layer4 = model.block3
    layer5 = nn.Sequential(model.bn1, model.relu)


.. GENERATED FROM PYTHON SOURCE LINES 52-53

Stage 2: Create and fit model

.. GENERATED FROM PYTHON SOURCE LINES 53-58

.. code-block:: Python
   :lineno-start: 53

    detector = MultiMahalanobis([layer1, layer2, layer3, layer4, layer5])

    print("Fitting...")
    detector.fit(train_loader, device=device)


.. GENERATED FROM PYTHON SOURCE LINES 59-60

Stage 3: Evaluate Detectors

.. GENERATED FROM PYTHON SOURCE LINES 60-68

.. code-block:: Python
   :lineno-start: 60

    print("Testing...")
    metrics = OODMetrics()

    for x, y in test_loader:
        metrics.update(detector(x.to(device)), y)

    print(metrics.compute())


.. GENERATED FROM PYTHON SOURCE LINES 69-71

This produces a table with the following output:
{'AUROC': 0.9601144790649414, 'AUPR-IN': 0.9439688324928284, 'AUPR-OUT': 0.9745389223098755, 'FPR95TPR': 0.23440000414848328}


.. _sphx_glr_download_auto_examples_detectors_mmahalanobis.py:

.. only:: html

  .. container:: sphx-glr-footer sphx-glr-footer-example

    .. container:: sphx-glr-download sphx-glr-download-jupyter

      :download:`Download Jupyter notebook: mmahalanobis.ipynb <mmahalanobis.ipynb>`

    .. container:: sphx-glr-download sphx-glr-download-python

      :download:`Download Python source code: mmahalanobis.py <mmahalanobis.py>`

    .. container:: sphx-glr-download sphx-glr-download-zip

      :download:`Download zipped: mmahalanobis.zip <mmahalanobis.zip>`


.. only:: html

 .. rst-class:: sphx-glr-signature

    `Gallery generated by Sphinx-Gallery <https://sphinx-gallery.github.io>`_