.. DO NOT EDIT.
.. THIS FILE WAS AUTOMATICALLY GENERATED BY SPHINX-GALLERY.
.. TO MAKE CHANGES, EDIT THE SOURCE PYTHON FILE:
.. "auto_examples/benchmarks/interface/cifar100_odin.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_benchmarks_interface_cifar100_odin.py>`
        to download the full example code.

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

.. _sphx_glr_auto_examples_benchmarks_interface_cifar100_odin.py:


ODIN - CIFAR100
==================

Reproduces the ODIN benchmark for OOD detection, from the paper
*Enhancing the reliability of out-of-distribution image detection in neural networks*.

.. GENERATED FROM PYTHON SOURCE LINES 10-24

.. code-block:: Python
   :lineno-start: 11


    import pandas as pd  # additional dependency, used here for convenience
    import torch

    from pytorch_ood.benchmark import CIFAR100_ODIN
    from pytorch_ood.detector import ODIN, MaxSoftmax
    from pytorch_ood.model import WideResNet
    from pytorch_ood.utils import fix_random_seed

    fix_random_seed(123)

    device = "cuda:0"
    loader_kwargs = {"batch_size": 64}


.. GENERATED FROM PYTHON SOURCE LINES 25-29

.. code-block:: Python
   :lineno-start: 25

    model = WideResNet(num_classes=100, pretrained="cifar100-pt").eval().to(device)
    trans = WideResNet.transform_for("cifar100-pt")
    norm_std = WideResNet.norm_std_for("cifar100-pt")


.. GENERATED FROM PYTHON SOURCE LINES 30-35

.. code-block:: Python
   :lineno-start: 30

    detectors = {
        "MSP": MaxSoftmax(model),
        "ODIN": ODIN(model, eps=0.002, norm_std=norm_std),
    }


.. GENERATED FROM PYTHON SOURCE LINES 36-50

.. code-block:: Python
   :lineno-start: 36

    results = []
    benchmark = CIFAR100_ODIN(root="data", transform=trans)

    with torch.no_grad():
        for detector_name, detector in detectors.items():
            print(f"> Evaluating {detector_name}")
            res = benchmark.evaluate(detector, loader_kwargs=loader_kwargs, device=device)
            for r in res:
                r.update({"Detector": detector_name})
            results += res

    df = pd.DataFrame(results)
    print((df.set_index(["Dataset", "Detector"]) * 100).to_csv(float_format="%.2f"))


.. GENERATED FROM PYTHON SOURCE LINES 51-80

This produces a table with the following output:

+--------------------+----------+-------+-------+---------+----------+----------+
| Dataset            | Detector | AUROC | AUTC  | AUPR-IN | AUPR-OUT | FPR95TPR |
+====================+==========+=======+=======+=========+==========+==========+
| TinyImageNetCrop   | MSP      | 86.32 | 31.64 | 88.23   | 84.81    | 43.36    |
+--------------------+----------+-------+-------+---------+----------+----------+
| TinyImageNetResize | MSP      | 74.64 | 40.18 | 77.29   | 70.91    | 65.56    |
+--------------------+----------+-------+-------+---------+----------+----------+
| LSUNResize         | MSP      | 75.38 | 39.77 | 78.50   | 71.16    | 63.36    |
+--------------------+----------+-------+-------+---------+----------+----------+
| LSUNCrop           | MSP      | 85.59 | 32.32 | 87.40   | 84.35    | 47.14    |
+--------------------+----------+-------+-------+---------+----------+----------+
| Uniform            | MSP      | 77.80 | 41.27 | 97.58   | 16.76    | 40.49    |
+--------------------+----------+-------+-------+---------+----------+----------+
| Gaussian           | MSP      | 84.97 | 35.02 | 98.43   | 23.36    | 29.45    |
+--------------------+----------+-------+-------+---------+----------+----------+
| TinyImageNetCrop   | ODIN     | 86.89 | 44.10 | 89.02   | 84.01    | 40.46    |
+--------------------+----------+-------+-------+---------+----------+----------+
| TinyImageNetResize | ODIN     | 80.79 | 44.87 | 82.08   | 78.43    | 60.08    |
+--------------------+----------+-------+-------+---------+----------+----------+
| LSUNResize         | ODIN     | 81.25 | 45.24 | 83.04   | 78.04    | 58.13    |
+--------------------+----------+-------+-------+---------+----------+----------+
| LSUNCrop           | ODIN     | 86.91 | 41.91 | 88.79   | 85.69    | 42.79    |
+--------------------+----------+-------+-------+---------+----------+----------+
| Uniform            | ODIN     | 95.42 | 35.01 | 99.54   | 58.03    | 14.21    |
+--------------------+----------+-------+-------+---------+----------+----------+
| Gaussian           | ODIN     | 98.51 | 24.27 | 99.85   | 84.87    | 5.76     |
+--------------------+----------+-------+-------+---------+----------+----------+


.. _sphx_glr_download_auto_examples_benchmarks_interface_cifar100_odin.py:

.. only:: html

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

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

      :download:`Download Jupyter notebook: cifar100_odin.ipynb <cifar100_odin.ipynb>`

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

      :download:`Download Python source code: cifar100_odin.py <cifar100_odin.py>`

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

      :download:`Download zipped: cifar100_odin.zip <cifar100_odin.zip>`


.. only:: html

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

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