.. DO NOT EDIT.
.. THIS FILE WAS AUTOMATICALLY GENERATED BY SPHINX-GALLERY.
.. TO MAKE CHANGES, EDIT THE SOURCE PYTHON FILE:
.. "auto_examples/self-supervised-learning/demo_poisson2sparse.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_self-supervised-learning_demo_poisson2sparse.py:


Poisson denoising using Poisson2Sparse
=======================================

This code shows how to restore a single image corrupted by Poisson noise using Poisson2Sparse, without requiring external training or knowledge of the noise level.

This method is based on the paper "Poisson2Sparse" :footcite:t:`ta2022poisson2sparse` and restores an image by learning a sparse non-linear dictionary parametrized by a neural network using a combination of Neighbor2Neighbor :footcite:t:`huang2021neighbor2neighbor`, of the negative log Poisson likelihood, of the :math:`\ell^1` pixel distance and of a sparsity-inducing :math:`\ell^1` regularization function on the weights.

.. GENERATED FROM PYTHON SOURCE LINES 10-15

.. code-block:: Python


    import deepinv as dinv
    import torch









.. GENERATED FROM PYTHON SOURCE LINES 16-20

Load a Poisson corrupted image
------------------------------

This example uses an image from the microscopy dataset FMD :footcite:t:`zhang2018poisson`.

.. GENERATED FROM PYTHON SOURCE LINES 20-36

.. code-block:: Python


    # Seed the RNGs for reproducibility
    torch.manual_seed(0)
    torch.cuda.manual_seed(0)

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

    physics = dinv.physics.Denoising(dinv.physics.PoissonNoise(gain=0.01, normalize=True))

    x = dinv.utils.demo.load_example(
        "FMD_TwoPhoton_MICE_R_gt_12_avg50.png", img_size=(256, 256)
    ).to(device)
    x = x[:, 0:1, :64, :64]
    x = x.clamp(0, 1)
    y = physics(x)








.. GENERATED FROM PYTHON SOURCE LINES 37-38

Define the Poisson2Sparse model

.. GENERATED FROM PYTHON SOURCE LINES 38-58

.. code-block:: Python


    backbone = dinv.models.ConvLista(
        in_channels=1,
        out_channels=1,
        kernel_size=3,
        num_filters=512,
        num_iter=10,
        stride=2,
        threshold=0.01,
    )

    model = dinv.models.Poisson2Sparse(
        backbone=backbone,
        lr=1e-4,
        num_iter=200,
        weight_n2n=2.0,
        weight_l1_regularization=1e-5,
        verbose=True,
    ).to(device)








.. GENERATED FROM PYTHON SOURCE LINES 59-64

Run the model
-------------

Note that we do not pass in the physics model as Poisson2Sparse assumes a
Poisson noise model internally and does not depend on the noise level.

.. GENERATED FROM PYTHON SOURCE LINES 64-80

.. code-block:: Python


    x_hat = model(y)

    # Compute and display PSNR values
    learning_free_psnr = dinv.metric.PSNR()(y, x).item()
    model_psnr = dinv.metric.PSNR()(x_hat, x).item()
    print(f"Measurement PSNR: {learning_free_psnr:.1f} dB")
    print(f"Poisson2Sparse PSNR: {model_psnr:.1f} dB")

    # Plot results
    dinv.utils.plot(
        [y, x_hat, x],
        titles=["Measurement", "Poisson2Sparse", "Ground truth"],
        subtitles=[f"{learning_free_psnr:.1f} dB", f"{model_psnr:.1f} dB", ""],
    )




.. image-sg:: /auto_examples/self-supervised-learning/images/sphx_glr_demo_poisson2sparse_001.png
   :alt: Measurement, Poisson2Sparse, Ground truth
   :srcset: /auto_examples/self-supervised-learning/images/sphx_glr_demo_poisson2sparse_001.png
   :class: sphx-glr-single-img


.. rst-class:: sphx-glr-script-out

 .. code-block:: none


      0%|          | 0/200 [00:00<?, ?it/s]
      0%|          | 1/200 [00:00<00:29,  6.64it/s]
      1%|          | 2/200 [00:00<00:28,  6.99it/s]
      2%|▏         | 3/200 [00:00<00:27,  7.12it/s]
      2%|▏         | 4/200 [00:00<00:26,  7.31it/s]
      2%|▎         | 5/200 [00:00<00:26,  7.42it/s]
      3%|▎         | 6/200 [00:00<00:25,  7.50it/s]
      4%|▎         | 7/200 [00:00<00:25,  7.55it/s]
      4%|▍         | 8/200 [00:01<00:25,  7.58it/s]
      4%|▍         | 9/200 [00:01<00:25,  7.58it/s]
      5%|▌         | 10/200 [00:01<00:25,  7.59it/s]
      6%|▌         | 11/200 [00:01<00:25,  7.55it/s]
      6%|▌         | 12/200 [00:01<00:25,  7.52it/s]
      6%|▋         | 13/200 [00:01<00:25,  7.44it/s]
      7%|▋         | 14/200 [00:01<00:25,  7.43it/s]
      8%|▊         | 15/200 [00:02<00:24,  7.47it/s]
      8%|▊         | 16/200 [00:02<00:24,  7.53it/s]
      8%|▊         | 17/200 [00:02<00:24,  7.41it/s]
      9%|▉         | 18/200 [00:02<00:24,  7.33it/s]
     10%|▉         | 19/200 [00:02<00:24,  7.35it/s]
     10%|█         | 20/200 [00:02<00:24,  7.34it/s]
     10%|█         | 21/200 [00:02<00:24,  7.32it/s]
     11%|█         | 22/200 [00:02<00:24,  7.29it/s]
     12%|█▏        | 23/200 [00:03<00:24,  7.36it/s]
     12%|█▏        | 24/200 [00:03<00:23,  7.41it/s]
     12%|█▎        | 25/200 [00:03<00:23,  7.42it/s]
     13%|█▎        | 26/200 [00:03<00:23,  7.43it/s]
     14%|█▎        | 27/200 [00:03<00:23,  7.42it/s]
     14%|█▍        | 28/200 [00:03<00:23,  7.42it/s]
     14%|█▍        | 29/200 [00:03<00:22,  7.46it/s]
     15%|█▌        | 30/200 [00:04<00:22,  7.48it/s]
     16%|█▌        | 31/200 [00:04<00:22,  7.47it/s]
     16%|█▌        | 32/200 [00:04<00:22,  7.33it/s]
     16%|█▋        | 33/200 [00:04<00:22,  7.36it/s]
     17%|█▋        | 34/200 [00:04<00:22,  7.46it/s]
     18%|█▊        | 35/200 [00:04<00:21,  7.54it/s]
     18%|█▊        | 36/200 [00:04<00:21,  7.51it/s]
     18%|█▊        | 37/200 [00:04<00:21,  7.43it/s]
     19%|█▉        | 38/200 [00:05<00:21,  7.44it/s]
     20%|█▉        | 39/200 [00:05<00:21,  7.46it/s]
     20%|██        | 40/200 [00:05<00:21,  7.53it/s]
     20%|██        | 41/200 [00:05<00:20,  7.60it/s]
     21%|██        | 42/200 [00:05<00:20,  7.64it/s]
     22%|██▏       | 43/200 [00:05<00:20,  7.60it/s]
     22%|██▏       | 44/200 [00:05<00:20,  7.63it/s]
     22%|██▎       | 45/200 [00:06<00:20,  7.62it/s]
     23%|██▎       | 46/200 [00:06<00:20,  7.64it/s]
     24%|██▎       | 47/200 [00:06<00:20,  7.55it/s]
     24%|██▍       | 48/200 [00:06<00:20,  7.58it/s]
     24%|██▍       | 49/200 [00:06<00:19,  7.62it/s]
     25%|██▌       | 50/200 [00:06<00:19,  7.55it/s]
     26%|██▌       | 51/200 [00:06<00:19,  7.51it/s]
     26%|██▌       | 52/200 [00:06<00:19,  7.47it/s]
     26%|██▋       | 53/200 [00:07<00:20,  7.33it/s]
     27%|██▋       | 54/200 [00:07<00:19,  7.36it/s]
     28%|██▊       | 55/200 [00:07<00:19,  7.36it/s]
     28%|██▊       | 56/200 [00:07<00:19,  7.42it/s]
     28%|██▊       | 57/200 [00:07<00:19,  7.25it/s]
     29%|██▉       | 58/200 [00:07<00:19,  7.24it/s]
     30%|██▉       | 59/200 [00:07<00:19,  7.25it/s]
     30%|███       | 60/200 [00:08<00:19,  7.26it/s]
     30%|███       | 61/200 [00:08<00:19,  7.27it/s]
     31%|███       | 62/200 [00:08<00:18,  7.28it/s]
     32%|███▏      | 63/200 [00:08<00:18,  7.32it/s]
     32%|███▏      | 64/200 [00:08<00:18,  7.37it/s]
     32%|███▎      | 65/200 [00:08<00:18,  7.49it/s]
     33%|███▎      | 66/200 [00:08<00:17,  7.56it/s]
     34%|███▎      | 67/200 [00:09<00:17,  7.60it/s]
     34%|███▍      | 68/200 [00:09<00:17,  7.57it/s]
     34%|███▍      | 69/200 [00:09<00:17,  7.51it/s]
     35%|███▌      | 70/200 [00:09<00:17,  7.54it/s]
     36%|███▌      | 71/200 [00:09<00:17,  7.57it/s]
     36%|███▌      | 72/200 [00:09<00:16,  7.61it/s]
     36%|███▋      | 73/200 [00:09<00:16,  7.65it/s]
     37%|███▋      | 74/200 [00:09<00:16,  7.62it/s]
     38%|███▊      | 75/200 [00:10<00:16,  7.59it/s]
     38%|███▊      | 76/200 [00:10<00:16,  7.66it/s]
     38%|███▊      | 77/200 [00:10<00:16,  7.66it/s]
     39%|███▉      | 78/200 [00:10<00:15,  7.66it/s]
     40%|███▉      | 79/200 [00:10<00:15,  7.63it/s]
     40%|████      | 80/200 [00:10<00:15,  7.59it/s]
     40%|████      | 81/200 [00:10<00:15,  7.57it/s]
     41%|████      | 82/200 [00:10<00:15,  7.49it/s]
     42%|████▏     | 83/200 [00:11<00:15,  7.57it/s]
     42%|████▏     | 84/200 [00:11<00:15,  7.62it/s]
     42%|████▎     | 85/200 [00:11<00:15,  7.66it/s]
     43%|████▎     | 86/200 [00:11<00:14,  7.65it/s]
     44%|████▎     | 87/200 [00:11<00:14,  7.62it/s]
     44%|████▍     | 88/200 [00:11<00:14,  7.63it/s]
     44%|████▍     | 89/200 [00:11<00:14,  7.61it/s]
     45%|████▌     | 90/200 [00:12<00:14,  7.54it/s]
     46%|████▌     | 91/200 [00:12<00:14,  7.48it/s]
     46%|████▌     | 92/200 [00:12<00:14,  7.46it/s]
     46%|████▋     | 93/200 [00:12<00:14,  7.51it/s]
     47%|████▋     | 94/200 [00:12<00:14,  7.50it/s]
     48%|████▊     | 95/200 [00:12<00:13,  7.57it/s]
     48%|████▊     | 96/200 [00:12<00:13,  7.60it/s]
     48%|████▊     | 97/200 [00:12<00:13,  7.60it/s]
     49%|████▉     | 98/200 [00:13<00:13,  7.62it/s]
     50%|████▉     | 99/200 [00:13<00:13,  7.56it/s]
     50%|█████     | 100/200 [00:13<00:13,  7.59it/s]
     50%|█████     | 101/200 [00:13<00:12,  7.63it/s]
     51%|█████     | 102/200 [00:13<00:12,  7.59it/s]
     52%|█████▏    | 103/200 [00:13<00:12,  7.60it/s]
     52%|█████▏    | 104/200 [00:13<00:12,  7.64it/s]
     52%|█████▎    | 105/200 [00:14<00:12,  7.64it/s]
     53%|█████▎    | 106/200 [00:14<00:12,  7.54it/s]
     54%|█████▎    | 107/200 [00:14<00:12,  7.58it/s]
     54%|█████▍    | 108/200 [00:14<00:12,  7.60it/s]
     55%|█████▍    | 109/200 [00:14<00:12,  7.57it/s]
     55%|█████▌    | 110/200 [00:14<00:11,  7.60it/s]
     56%|█████▌    | 111/200 [00:14<00:11,  7.58it/s]
     56%|█████▌    | 112/200 [00:14<00:11,  7.65it/s]
     56%|█████▋    | 113/200 [00:15<00:11,  7.67it/s]
     57%|█████▋    | 114/200 [00:15<00:11,  7.70it/s]
     57%|█████▊    | 115/200 [00:15<00:11,  7.68it/s]
     58%|█████▊    | 116/200 [00:15<00:10,  7.70it/s]
     58%|█████▊    | 117/200 [00:15<00:10,  7.71it/s]
     59%|█████▉    | 118/200 [00:15<00:10,  7.68it/s]
     60%|█████▉    | 119/200 [00:15<00:10,  7.67it/s]
     60%|██████    | 120/200 [00:15<00:10,  7.61it/s]
     60%|██████    | 121/200 [00:16<00:10,  7.66it/s]
     61%|██████    | 122/200 [00:16<00:10,  7.68it/s]
     62%|██████▏   | 123/200 [00:16<00:10,  7.65it/s]
     62%|██████▏   | 124/200 [00:16<00:09,  7.67it/s]
     62%|██████▎   | 125/200 [00:16<00:09,  7.58it/s]
     63%|██████▎   | 126/200 [00:16<00:10,  7.37it/s]
     64%|██████▎   | 127/200 [00:16<00:10,  7.27it/s]
     64%|██████▍   | 128/200 [00:17<00:09,  7.27it/s]
     64%|██████▍   | 129/200 [00:17<00:09,  7.25it/s]
     65%|██████▌   | 130/200 [00:17<00:09,  7.17it/s]
     66%|██████▌   | 131/200 [00:17<00:09,  7.22it/s]
     66%|██████▌   | 132/200 [00:17<00:09,  7.21it/s]
     66%|██████▋   | 133/200 [00:17<00:09,  7.24it/s]
     67%|██████▋   | 134/200 [00:17<00:09,  7.28it/s]
     68%|██████▊   | 135/200 [00:18<00:08,  7.31it/s]
     68%|██████▊   | 136/200 [00:18<00:08,  7.37it/s]
     68%|██████▊   | 137/200 [00:18<00:08,  7.46it/s]
     69%|██████▉   | 138/200 [00:18<00:08,  7.55it/s]
     70%|██████▉   | 139/200 [00:18<00:08,  7.58it/s]
     70%|███████   | 140/200 [00:18<00:07,  7.53it/s]
     70%|███████   | 141/200 [00:18<00:07,  7.51it/s]
     71%|███████   | 142/200 [00:18<00:07,  7.49it/s]
     72%|███████▏  | 143/200 [00:19<00:07,  7.51it/s]
     72%|███████▏  | 144/200 [00:19<00:07,  7.48it/s]
     72%|███████▎  | 145/200 [00:19<00:07,  7.50it/s]
     73%|███████▎  | 146/200 [00:19<00:07,  7.53it/s]
     74%|███████▎  | 147/200 [00:19<00:07,  7.50it/s]
     74%|███████▍  | 148/200 [00:19<00:06,  7.56it/s]
     74%|███████▍  | 149/200 [00:19<00:06,  7.65it/s]
     75%|███████▌  | 150/200 [00:19<00:06,  7.73it/s]
     76%|███████▌  | 151/200 [00:20<00:06,  7.76it/s]
     76%|███████▌  | 152/200 [00:20<00:06,  7.78it/s]
     76%|███████▋  | 153/200 [00:20<00:06,  7.78it/s]
     77%|███████▋  | 154/200 [00:20<00:05,  7.80it/s]
     78%|███████▊  | 155/200 [00:20<00:05,  7.76it/s]
     78%|███████▊  | 156/200 [00:20<00:05,  7.71it/s]
     78%|███████▊  | 157/200 [00:20<00:05,  7.62it/s]
     79%|███████▉  | 158/200 [00:21<00:05,  7.58it/s]
     80%|███████▉  | 159/200 [00:21<00:05,  7.63it/s]
     80%|████████  | 160/200 [00:21<00:05,  7.66it/s]
     80%|████████  | 161/200 [00:21<00:05,  7.71it/s]
     81%|████████  | 162/200 [00:21<00:04,  7.70it/s]
     82%|████████▏ | 163/200 [00:21<00:04,  7.70it/s]
     82%|████████▏ | 164/200 [00:21<00:04,  7.76it/s]
     82%|████████▎ | 165/200 [00:21<00:04,  7.80it/s]
     83%|████████▎ | 166/200 [00:22<00:04,  7.76it/s]
     84%|████████▎ | 167/200 [00:22<00:04,  7.72it/s]
     84%|████████▍ | 168/200 [00:22<00:04,  7.64it/s]
     84%|████████▍ | 169/200 [00:22<00:04,  7.54it/s]
     85%|████████▌ | 170/200 [00:22<00:03,  7.54it/s]
     86%|████████▌ | 171/200 [00:22<00:03,  7.53it/s]
     86%|████████▌ | 172/200 [00:22<00:03,  7.50it/s]
     86%|████████▋ | 173/200 [00:23<00:03,  7.42it/s]
     87%|████████▋ | 174/200 [00:23<00:03,  7.10it/s]
     88%|████████▊ | 175/200 [00:23<00:03,  6.93it/s]
     88%|████████▊ | 176/200 [00:23<00:03,  7.03it/s]
     88%|████████▊ | 177/200 [00:23<00:03,  7.08it/s]
     89%|████████▉ | 178/200 [00:23<00:03,  7.13it/s]
     90%|████████▉ | 179/200 [00:23<00:02,  7.18it/s]
     90%|█████████ | 180/200 [00:23<00:02,  7.33it/s]
     90%|█████████ | 181/200 [00:24<00:02,  7.36it/s]
     91%|█████████ | 182/200 [00:24<00:02,  7.30it/s]
     92%|█████████▏| 183/200 [00:24<00:02,  7.34it/s]
     92%|█████████▏| 184/200 [00:24<00:02,  7.38it/s]
     92%|█████████▎| 185/200 [00:24<00:02,  7.37it/s]
     93%|█████████▎| 186/200 [00:24<00:01,  7.53it/s]
     94%|█████████▎| 187/200 [00:24<00:01,  7.61it/s]
     94%|█████████▍| 188/200 [00:25<00:01,  7.61it/s]
     94%|█████████▍| 189/200 [00:25<00:01,  7.65it/s]
     95%|█████████▌| 190/200 [00:25<00:01,  7.70it/s]
     96%|█████████▌| 191/200 [00:25<00:01,  7.52it/s]
     96%|█████████▌| 192/200 [00:25<00:01,  7.34it/s]
     96%|█████████▋| 193/200 [00:25<00:00,  7.29it/s]
     97%|█████████▋| 194/200 [00:25<00:00,  7.31it/s]
     98%|█████████▊| 195/200 [00:26<00:00,  7.29it/s]
     98%|█████████▊| 196/200 [00:26<00:00,  7.13it/s]
     98%|█████████▊| 197/200 [00:26<00:00,  7.22it/s]
     99%|█████████▉| 198/200 [00:26<00:00,  7.29it/s]
    100%|█████████▉| 199/200 [00:26<00:00,  7.35it/s]
    100%|██████████| 200/200 [00:26<00:00,  7.45it/s]
    100%|██████████| 200/200 [00:26<00:00,  7.49it/s]
    Measurement PSNR: 27.3 dB
    Poisson2Sparse PSNR: 31.1 dB




.. GENERATED FROM PYTHON SOURCE LINES 81-84

:References:

.. footbibliography::


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

   **Total running time of the script:** (0 minutes 27.220 seconds)


.. _sphx_glr_download_auto_examples_self-supervised-learning_demo_poisson2sparse.py:

.. only:: html

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

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

      :download:`Download Jupyter notebook: demo_poisson2sparse.ipynb <demo_poisson2sparse.ipynb>`

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

      :download:`Download Python source code: demo_poisson2sparse.py <demo_poisson2sparse.py>`

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

      :download:`Download zipped: demo_poisson2sparse.zip <demo_poisson2sparse.zip>`


.. only:: html

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

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