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

.. only:: html

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

        New to DeepInverse? Get started with the basics with the
        :ref:`5 minute quickstart tutorial <sphx_glr_auto_examples_basics_demo_quickstart.py>`.

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

.. _sphx_glr_auto_examples_basics_demo_pretrained_model.py:


Use a pretrained model
====================================================================================================

Follow this example to reconstruct images using a pretrained model in one line.

We show three sets of general pretrained reconstruction methods, including:

* Pretrained feedforward :class:`Reconstruct Anything Model (RAM) <deepinv.models.RAM>`;
* :ref:`Plug-and-play <iterative>` with a pretrained denoiser.
* Pretrained :ref:`diffusion model <diffusion>`;

See :ref:`pretrained models <pretrained-models>` for a principled comparison between methods demonstrated in this example.

.. tip::

    * Want to use your own dataset? See :ref:`sphx_glr_auto_examples_basics_demo_custom_dataset.py`
    * Want to use your own physics? See :ref:`sphx_glr_auto_examples_basics_demo_custom_physics.py`

.. GENERATED FROM PYTHON SOURCE LINES 21-27

.. code-block:: Python


    import deepinv as dinv
    import torch

    device = dinv.utils.get_freer_gpu() if torch.cuda.is_available() else "cpu"








.. GENERATED FROM PYTHON SOURCE LINES 28-29

Let's say you want to reconstruct a butterfly from noisy, blurry measurements:

.. GENERATED FROM PYTHON SOURCE LINES 29-43

.. code-block:: Python


    # Ground truth
    x = dinv.utils.load_example("butterfly.png", device=device)

    # Define physics
    physics = dinv.physics.BlurFFT(
        x.shape[1:],
        filter=dinv.physics.blur.gaussian_blur((5, 5)),
        noise_model=dinv.physics.GaussianNoise(sigma=0.1),
        device=device,
    )

    y = physics(x)








.. GENERATED FROM PYTHON SOURCE LINES 44-49

For each model, define model in one line and reconstruct in one line.
Pretrained Reconstruct Anything Model:

.. seealso::
    See :ref:`sphx_glr_auto_examples_models_demo_foundation_model.py` for further one-line examples for the RAM model across various domains.

.. GENERATED FROM PYTHON SOURCE LINES 49-55

.. code-block:: Python


    model = dinv.models.RAM(pretrained=True, device=device)

    with torch.no_grad():
        x_hat1 = model(y, physics)








.. GENERATED FROM PYTHON SOURCE LINES 56-62

PnP algorithm with pretrained denoiser:

.. seealso::
    See :ref:`pretrained denoisers <pretrained-weights>` for a full list of denoisers that can be plugged into iterative/sampling algorithms.

    See :ref:`sphx_glr_auto_examples_plug-and-play_demo_PnP_DPIR_deblur.py` for a further example of using plug-and-play with a pretrained denoiser.

.. GENERATED FROM PYTHON SOURCE LINES 62-68

.. code-block:: Python


    denoiser = dinv.models.DRUNet(device=device)
    model = dinv.optim.DPIR(sigma=0.1, denoiser=denoiser, device=device)

    x_hat2 = model(y, physics)








.. GENERATED FROM PYTHON SOURCE LINES 69-73

Pretrained diffusion model (we reduce the image size for demo speed on CPU, as diffusion model is slow):

.. seealso::
    See :ref:`sphx_glr_auto_examples_sampling_demo_ddrm.py` for a further example of using a pretrained diffusion model.

.. GENERATED FROM PYTHON SOURCE LINES 73-79

.. code-block:: Python



    model = dinv.sampling.DDRM(denoiser, sigmas=torch.linspace(1, 0, 20)).to(device)

    x_hat3 = model(y, physics)








.. GENERATED FROM PYTHON SOURCE LINES 80-81

Plot results

.. GENERATED FROM PYTHON SOURCE LINES 81-92

.. code-block:: Python

    dinv.utils.plot(
        {
            "Ground truth": x,
            "Blurred measurement": y,
            "Pretrained RAM": x_hat1,
            "Pretrained PnP": x_hat2,
            "Pretrained diffusion": x_hat3,
        }
    )





.. image-sg:: /auto_examples/basics/images/sphx_glr_demo_pretrained_model_001.png
   :alt: Ground truth, Blurred measurement, Pretrained RAM, Pretrained PnP, Pretrained diffusion
   :srcset: /auto_examples/basics/images/sphx_glr_demo_pretrained_model_001.png
   :class: sphx-glr-single-img





.. GENERATED FROM PYTHON SOURCE LINES 93-100

🎉 Well done, you now know how to use pretrained models!

What's next?
~~~~~~~~~~~~
* Check out the :ref:`example on how to fine-tune a foundation model to your own problem <sphx_glr_auto_examples_models_demo_foundation_model.py>`.
* See :ref:`pretrained models <pretrained-models>` for a comparison between methods demonstrated in this example.
* See :ref:`diffusion <diffusion>` and :ref:`iterative <iterative>` for how to fully customize your sampling or iterative algorithm using a pretrained denoiser.


.. rst-class:: sphx-glr-timing

   **Total running time of the script:** (0 minutes 50.291 seconds)


.. _sphx_glr_download_auto_examples_basics_demo_pretrained_model.py:

.. only:: html

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

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

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

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

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

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

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


.. only:: html

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

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