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

.. only:: html

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

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

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

.. _sphx_glr_csi_Reconstruction_tutorial.py:


CSI reconstruction Tutorial
==============================================================

.. GENERATED FROM PYTHON SOURCE LINES 7-10

.. code-block:: Python


    # !pip install pysensing








.. GENERATED FROM PYTHON SOURCE LINES 11-13

In this tutorial, we will be implementing codes for CSI human pose estimation task


.. GENERATED FROM PYTHON SOURCE LINES 13-22

.. code-block:: Python

    import sys
    sys.path.append('../..')
    import pysensing.csi.dataset.get_dataloader as get_dataloader
    import pysensing.csi.model.get_model as get_model
    import pysensing.csi.inference.predict as predict
    import pysensing.csi.inference.train as train
    import pysensing.csi.inference.embedding as embedding
    import torch








.. GENERATED FROM PYTHON SOURCE LINES 23-26

Load the data
-----------------------------------
CSI reconstruction dataset: 

.. GENERATED FROM PYTHON SOURCE LINES 26-52

.. code-block:: Python


    # HandFi
    # CSI size : 6, 20, 114
    # image : 144, 144
    # joints2d :  2, 42
    # joints3d : 2, 21
    # train number : 3600
    # test number : 400


    train_loader, test_loader = get_dataloader.load_recon_dataset('HandFi', batch_size=32, return_train=True)
    device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
    for data in train_loader:
        ((joints,image), csi) = data
        joint = joints[:,:,0:21].to(device,dtype=torch.float)
        img=image.to(device,dtype=torch.float)
        csi=csi.to(device,dtype=torch.float)
        joint2d = joint[:,0:2,:] 
        joint2d = joint2d.view(-1,42)
        joint3d = joint[:,2,:] 
    
        print('data:', csi)
        print('img:', img)
        print('joint:', joint)
        break



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

.. code-block:: pytb

    Traceback (most recent call last):
      File "/data1/msc/zyj/yunjiao_csi/1028/yunjiao_csi/tutorials/csi_source/Reconstruction_tutorial.py", line 36, in <module>
        train_loader, test_loader = get_dataloader.load_recon_dataset('HandFi', batch_size=32, return_train=True)
                                    ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
      File "/data1/msc/zyj/yunjiao_csi/1028/yunjiao_csi/tutorials/csi_source/../../pysensing/csi/dataset/get_dataloader.py", line 200, in load_recon_dataset
        raise ValueError("Download failed.")
    ValueError: Download failed.




.. GENERATED FROM PYTHON SOURCE LINES 53-56

Load the model
-----------------------------------
For HandFi dataset, model zoo contains AutoEncoder.

.. GENERATED FROM PYTHON SOURCE LINES 56-60

.. code-block:: Python


    model = get_model.load_recon_model('HandFi', 'AutoEncoder')
    print(model)


.. GENERATED FROM PYTHON SOURCE LINES 61-63

Model train
------------------------

.. GENERATED FROM PYTHON SOURCE LINES 63-68

.. code-block:: Python

    optimizer = torch.optim.Adam(model.parameters(), lr=0.001)
    epoch_num = 1

    train.recon_train(train_loader, model, epoch_num, optimizer, device)


.. GENERATED FROM PYTHON SOURCE LINES 69-71

Model inference
------------------------

.. GENERATED FROM PYTHON SOURCE LINES 71-80

.. code-block:: Python


    model = get_model.load_pretrain(model, 'HandFi', 'AutoEncoder', device=device)
    output = predict.recon_predict(csi, 'HandFi', model, device)
    _, mask, twod, threed = output 
    print("mask:", mask.shape)
    print("twod:", twod.shape)
    print("threed:", threed.shape)



.. GENERATED FROM PYTHON SOURCE LINES 81-83

Evaluate the loss
------------------------

.. GENERATED FROM PYTHON SOURCE LINES 83-89

.. code-block:: Python


    IoUerr = train.IoU(img,mask) 
    mPAerr = train.mPA(img,mask)
    mpjpe, pck = train.mpjpe_pck(joint2d,joint3d, twod, threed)

    print(  f'mPA: {mPAerr:.3f} | => IoU: {IoUerr:.3f} | => mpjpe: {mpjpe:.3f} | =>pck: {pck:.3f}\n')

.. GENERATED FROM PYTHON SOURCE LINES 90-92

Generate embedding
------------------------

.. GENERATED FROM PYTHON SOURCE LINES 92-97

.. code-block:: Python


    csi_embedding = embedding.recon_csi_embedding(csi, 'HandFi', model, device)
    print('csi_embedding: ', csi_embedding)



.. GENERATED FROM PYTHON SOURCE LINES 98-99

And that's it. We're done with our CSI reconstruction tutorials. Thanks for reading.


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

   **Total running time of the script:** (0 minutes 0.592 seconds)


.. _sphx_glr_download_csi_Reconstruction_tutorial.py:

.. only:: html

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

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

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

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

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


.. only:: html

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

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